{ "id": 321733, "title": "Papers on Plant Identification Deep Learning Techniques", "url": "/competitions/herbarium-2022-fgvc9/discussion/321733", "author_name": "", "post_date": "2022-04-28T10:59:35.111000", "votes": 5, "comment_count": 1, "views": 0, "content": "

These are the papers that will change your life.
\nThese are the papers that will make you question everything you know.
\nThese are the papers that will make you see the world in a whole new way.
\nThese are the papers that will give you hints on improving your LB score ;)
\nHave fun!

\n
\n

Medicinal Plant Identification using Gabor Filters and Deep Learning Techniques: A Paper Review

\n
\n

TL;DR: Computer-aided identification of plants is a branch of machine learning that has become more recognized recently and proves itself as a vital tool in numerous sectors including pharmacological science, forestry and agriculture. This study reviewed plant species classification relying on leaf textural features using Gabor filters and revealed that Gabor filters perform better when combined with other feature extraction methods. Therefore, this study proposes using Log-Gabor filter in the field of plant identification to improve accuracy since they overcome the drawbacks of Gabor filters which are; the maximum bandwidth of a Gabor filter is limited to approximately one octave and Gabor filters are not optimal if one
\n Corresponding Author: Stephen Opoku Oppong ICT Education Department, University of Education, Winneba, Ghana Email: sooppong@uew.edu.gh Abstract: Computer-aided identification of plants is a branch of machine learning that has become more recognized recently and proves itself as a vital tool in numerous sectors including pharmacological science, forestry and agriculture. This has essentially generated a zeal in creating automated systems for the identification of diverse species of plants. This study reviewed plant species classification relying on leaf textural features using Gabor filters and revealed that Gabor filters perform better when combined with other feature extraction methods. Therefore, this study proposes using Log-Gabor filter in the field of plant identification to improve accuracy since they overcome the drawbacks of Gabor filters which are; the maximum bandwidth of a Gabor filter is limited to approximately one octave and Gabor filters are not optimal if one is seeking broad spectral information with maximal spatial localization.

\n
\n
\n

Holistic Based Plant Identification Using Deep Learning
\n TL;DR: This paper introduces an automated plant classification framework for recognizing plants dependent on their holistic shape. The system uses image preprocessing, deep learning techniques, and ResNet18 to identify plants through the camera; it will give their names in Urdu, English, and the plant's scientific name as an output. Moreover, the system can describe the plant along with its audio. The methodology was evaluated using two datasets. The first dataset, GRANDYMU, consists of 70 different plant species containing 3500 images, and the second is the famous SWEDISH leaf dataset. We trained different models based on Resnet18
\n Plants are the most important forms of life on earth. They are the fundamental asset for human prosperity, furnishing us with oxygen and food. Humans have widely utilized plants in medication, food formation, and the cosmetic business worldwide in their daily life activities. Identification of plants is extremely a challenging task due to the recursive nature of the shapes of plants. Moreover, identification has been done on a few categories and is limited to veins or leaf patterns. This paper introduced an automated plant classification framework for recognizing plants dependent on their holistic shape. The system uses image preprocessing, deep learning techniques, and ResNet18 to identify plants through the camera; it will give their names in Urdu, English, and the plant's scientific name as an output. Moreover, the system can describe the plant along with its audio. The methodology was evaluated using two datasets. The first dataset, GRANDYMU, consists of 70 different plant species containing 3500 images, and the second is the famous SWEDISH leaf dataset. We trained different models based on Resnet18, and the maximum accuracy that we achieved was 99%. Also, we got an accuracy of 99.9% for the latter dataset.

\n
\n
\n

Experimental evaluation of Data Augmentation heuristics for plant identification systems based on Deep Learning
\n Data augmentation (DA) allows increasing datasets for training machine learning models that demands large amounts of data. In real-world applications in which data may not be abundant enough and data acquisition is not easy, DA enables increasing diversity and introducing model generalization. In this work we evaluate several DA techniques and combining approaches to extend image datasets used to train plant species recognition models. We experimentally validated Deep Convolutional Neural Networks (DCNN) with several datasets obtained from common augmentation techniques and combinations. The results allowed the identification of the Translate + Crop augmentation policy as the most effective within the scope of evaluation.

\n
\n
\n

Deep Learning for Plant Identification in Natural Environment
\n TL;DR: We collected a dataset of 10,000 images of 100 ornamental plant species in Beijing Forestry University campus using mobile phones. A 26-layer deep learning model consisting of 8 residual building blocks was designed for large-scale plant classification in natural environment. The proposed model achieved a recognition rate of 91.78% on the BJFU100 dataset, demonstrating that deep learning is a promising technology for smart forestry.
\n Plant image identification has become an interdisciplinary focus in both botanical taxonomy and computer vision. The first plant image dataset collected by mobile phone in natural scene is presented, which contains 10,000 images of 100 ornamental plant species in Beijing Forestry University campus. A 26-layer deep learning model consisting of 8 residual building blocks is designed for large-scale plant classification in natural environment. The proposed model achieves a recognition rate of 91.78% on the BJFU100 dataset, demonstrating that deep learning is a promising technology for smart forestry.

\n
\n
\n

Plant Identification Based on Noisy Web Data: the Amazing Performance of Deep Learning (LifeCLEF 2017)
\n The 2017 fith edition of the LifeCLEF plant identification challenge is an important milestone towards automated plant identification systems working at the scale of continental floras with 10.000 plant species living mainly in Europe and North America illustrated by a total of 1.1M images. Nowadays, such ambitious systems are enabled thanks to the conjunction of the dazzling recent progress in image classification with deep learning and several outstanding international initiatives, such as the Encyclopedia of Life (EOL), aggregating the visual knowledge on plant species coming from the main national botany institutes. However, despite all these efforts the majority of the plant species still remain without pictures or are poorly illustrated. Outside the institutional channels, a much larger number of plant pictures are available and spread on the web through botanist blogs, plant lovers web-pages, image hosting websites and on-line plant retailers. The LifeCLEF 2017 plant challenge presented in this paper aimed at evaluating to what extent a large noisy training dataset collected through the web and containing a lot of labelling errors can compete with a smaller but trusted training dataset checked by experts. To fairly compare both training strategies, the test dataset was created from a third data source, i.e. the Pl@ntNet mobile application that collects millions of plant image queries all over the world. This paper presents more precisely the resources and assessments of the challenge, summarizes the approaches and systems employed by the participating research groups, and provides an analysis of the main outcomes. (Resume d'auteur)

\n
\n
\n

Deep Learning with Taxonomic Loss for Plant Identification
\n TL;DR: Train a neural network on a plant identification dataset using a taxonomic loss function to improve performance.
\n Plant identification is a fine-grained classification task which aims to identify the family, genus, and species according to plant appearance features. Inspired by the hierarchical structure of taxonomic tree, the taxonomic loss was proposed, which could encode the hierarchical relationships among multilevel labels into the deep learning objective function by simple group and sum operation. By training various neural networks on PlantCLEF 2015 and PlantCLEF 2017 datasets, the experimental results demonstrated that the proposed loss function was easy to implement and outperformed the most commonly adopted cross-entropy loss. Eight neural networks were trained, respectively, by two different loss functions on PlantCLEF 2015 dataset, and the models trained by taxonomic loss led to significant performance improvements. On PlantCLEF 2017 dataset with 10,000 species, the SENet-154 model trained by taxonomic loss achieved the accuracies of 84.07%, 79.97%, and 73.61% at family, genus and species levels, which improved those of model trained by cross-entropy loss by 2.23%, 1.34%, and 1.08%, respectively. The taxonomic loss could further facilitate the fine-grained classification task with hierarchical labels.

\n
\n
\n

A Light Weight Deep Learning Model for Real World Plant Identification
\n Automatic identification and classification of different plant leaf species have become a common trend among researchers and scientists. To obtain a result with better precision, they use various methods and techniques of deep learning to build a model. Convolutional neural networks are becoming the most common method used by scientists to classify plant leaves. However, the classification of plant leaves can be challenging with more rare species and complicated backgrounds, for which researchers build several models to achieve high-level accuracy. In the present study for the classification of leaves, we have created a model for plant leaf classification based on a dataset we collected. We've used the Resnet-50 model, a well-known CNN architecture, which provided an efficient method to organize and analyze a deep classification to reduce the complexity so that there will be fewer parameters for training and low time consumption as well. Using Resnet-50, we intended to develop a significant result in our classification model. The convolutional neural network is famous for its influential abilities in feature extraction and classification. And Resnet-50 being a residual network enabled us to train deep networks in our model. The average training accuracy reached 98.3%, while the average testing accuracy reached 92.5%. The key contribution of this study is effective accuracy as well as we have trained the model on our own prepared dataset that we have prepared from real world environment. Data Availability: https://drive.google.com/file/d/1bD7B257l-6wqUCQHBWhle95xyrotUbwO/view?usp=sharing

\n
\n
\n

Fine-Grained Plant Identification using wide and deep learning model 1
\n TL;DR: We propose a model to address the fine-grained plant image classification task by using the wide and deep learning framework which combines a linear model and a deep learning model. Proposed method sums the result of the wide and deep learning model using a logistic function so that discrete features can be considered simultaneously with continuous image content.
\n In recent years, with the evolution of deep learning technology, the performance of plant image recognition has improved remarkably. In this paper, we propose a model to address the fine-grained plant image classification task by using the wide and deep learning framework which combines a linear model and a deep learning model. Proposed method sums the result of the wide and deep learning model using a logistic function so that discrete features can be considered simultaneously with continuous image content. Our works used metadata such as the date of flowering and locational information for the wide model. Our experiment shows that the proposed method gives better performance than a baseline method.

\n
\n
\n

Plant Identification: Experts vs. Machines in the Era of Deep Learning - Deep Learning Techniques Challenge Flora Experts
\n TL;DR: Nine deep learning models were evaluated with regard to nine French botanists. The performance of the deep learning models was found to be close to the human expertise.
\n Automated identification of plants and animals have improved considerably in the last few years, in particular thanks to the recent advances in deep learning. The next big question is how far such automated systems are from the human expertise. Indeed, even the best experts are sometimes confused and/or disagree between each others when validating visual or audio observations of living organism. A picture or a sound actually contains only a partial information that is usually not sufficient to determine the right species with certainty. Quantifying this uncertainty and comparing it to the performance of automated systems is of high interest for both computer scientists and expert naturalists. This chapter reports an experimental study following this idea in the plant domain. In total, nine deep-learning systems implemented by three different research teams were evaluated with regard to nine expert botanists of the French flora. Therefore, we created a small set of plant observations that were identified in the field and revised by experts in order to have a near-perfect golden standard. The main outcome of this work is that the performance of state-of-the-art deep learning models is now close to the most advanced human expertise. This shows that automated plant identification systems are now mature enough for several routine tasks, and can offer very promising tools for autonomous ecological surveillance systems.

\n
\n
\n

Observation on Therapeutic Plant Identification based on Deep Learning Technique
\n TL;DR: There are many different methods for leaf classification, but the most successful ones rely on deep learning algorithms. These algorithms can learn to identify patterns in data that are too complex for humans to discern. This makes them very effective at identifying different types of leaves.
\n Plants have been used for medicinal purposes long before recorded history. It plays a major role in medicines, food, perfumes and cosmetics industries. By knowing the herbal plants and its usage it can be used for above applications. In this digital era, people don’t have adequate knowledge to identify various herbal plants which are used by our ancestors for long time. Presently, the identification of herbal plants is purely based on the human perception or knowledge. There may be probability of human error occurring. In order to have an efficient herb species classification, there must be a complete model which should be automatic and convenient recognition system. This paper is reviewing the different leaf classification methodologies based on deep learning algorithms. The main aim of this research paper is to conclude the advanced technique for the leaf identification.

\n
\n
\n

Deep learning for plant identification: how the web can compete with human experts
\n Automated identification of plants and animals has improved considerably in the last few years, in particular thanks to the recent advances in deep learning. In order to evaluate the performance of automated plant identification technologies in a sustainable and repeatable way, a dedicated system-oriented benchmark was setup in 2011 in the context of ImageCLEF (Goëau et al. 2011). Each year, since that time, several research groups participated in this large collaborative evaluation by benchmarking their image-based plant identification systems. In 2014, the LifeCLEF research platform (Joly et al. 2014) was created in the continuity of this effort so as to enlarge the evaluated challenges by considering birds and fishes in addition to plants, and audio and video contents in addition to images. The 2017-th edition of the LifeCLEF plant identification challenge (Joly et al. 2017) is an important milestone towards automated plant identification systems working at the scale of continental floras with 10.000 plant species living mainly in Europe and North America ‡,§ | §,‡ ¶ # ¤ © Goëau H et al. This is an open access article distributed under the terms of the Creative Commons Attribution License (CC BY 4.0), which permits unrestricted use, distribution, and reproduction in any medium, provided the original author and source are credited. illustrated by a total of 1.1M images. Nowadays, such ambitious systems are enabled thanks to the conjunction of the dazzling recent progress in image classification with deep learning and several outstanding international initiatives, aggregating the visual knowledge on plant species coming from the main national botanical institutes. The PlantCLEF plant challenge that we propose to present at this workshop aimed at evaluating to what extent a large noisy training dataset collected through the web (then containing a lot of labelling errors) can compete with a smaller but trusted training dataset checked by experts. To fairly compare both training strategies, the test dataset was created from a third data source, the Pl@ntNet (Joly et al. 2015) mobile application that collects millions of plant image queries all over the world. Due to the good results obtained at the 2017-th edition of the LifeCLEF plant identification challenge, the next big question is how far such automated systems are from the human expertise. Indeed, even the best experts are sometimes confused and/or disagree with each other when validating images of living organism. A multimedia data actually contains only partial information that is usually not sufficient to determine the right species with certainty. Quantifying this uncertainty and comparing it to the performance of automated systems is of high interest for both computer scientists and expert naturalists. This work reports an experimental study following this idea in the plant domain. In total, 9 deeplearning systems implemented by 3 different research teams were evaluated with regard to 9 expert botanists of the French flora. The main outcome of this work is that the performance of state-of-the-art deep learning models is now close to the most advanced human expertise. This shows that automated plant identification systems are now mature enough for several routine tasks, and can offer very promising tools for autonomous ecological surveillance systems.

\n
\n
\n

Plant Identification with Deep Learning Ensembles
\n This work describes the plant identification system that we submitted to the ExpertLifeCLEF plant identification campaign in 2018. We fine-tuned two pre-trained deep learning architectures (SeNet and DensNetwork) using images shared by the CLEF organizers in 2017. Our main runs are 4 ensembles obtained with different weighted combinations of the 4 deep learning architectures. The fifth ensemble is based on deep learning features but uses Error Correcting Output Codes (ECOC) as the ensemble. Our best system has achieved a classification accuracy of 74.4%, while the best system obtained 86.7% accuracy, on the whole of the official test data. This system ranked 4th place among all the teams, but matched the accuracy of one of the human experts.

\n
\n
\n

A Combination of Deep Learning and Hand-Designed Feature for Plant Identification Based on Leaf and Flower Images
\n TL;DR: This paper proposes a combination of deep learning and hand-designed feature for plant identification based on leaf and flower images. The contributions of this paper are two-fold. First, for each organ image, we have performed a comparative evaluation of deep learning and hand-designed feature for plant identification. Two approaches for deep learning and hand-designed feature that are convolutional neuron network (CNN) and kernel descriptor (KDES) are chosen in our experiments. Second, based on the results of the first contribution, we propose a method for plant identification by late fusing the identification results of leaf and flower. Experimental results on ImageCle
\n This paper proposes a combination of deep learning and hand-designed feature for plant identification based on leaf and flower images. The contributions of this paper are two-fold. First, for each organ image, we have performed a comparative evaluation of deep learning and hand-designed feature for plant identification. Two approaches for deep learning and hand-designed feature that are convolutional neuron network (CNN) and kernel descriptor (KDES) are chosen in our experiments. Second, based on the results of the first contribution, we propose a method for plant identification by late fusing the identification results of leaf and flower. Experimental results on ImageClef 2015 dataset show that hand designed feature outperforms deep learning for well-constrained cases (leaf captured on simple background). However, deep learning shows its robustness in natural situations. Moreover, the combination of leaf and flower images improves significantly the identification when comparing leaf-based plant identification.

\n
\n
\n

Transfer learning with deep convolutional neural network for automated plant identification in hainan island
\n The Hainan Island has a generally high biological diversity with a wide variety of plant species, some of which are listed as endemic to the island. It is time-consuming and difficult, even for the botanist experts to determine the name of species based on observations. Automated plant identification enables experts to process significantly greater numbers of plants with higher efficiencies in shorter periods of time. However, plant recognition is a kind of fine-grained visual recognition problem, which is relatively harder than conventional image recognition. In this paper, we employ a Deep Convolutional Neural Network (DCNN) trained on the ImageNet database, which contains millions of images, and then transfer the learning information for automated plant identification based on flower and fruit images. First, we modify the last three layers of the pre-trained network in order to adapt ResNet-50 model to our classification task, and replace the fully connected layer in the original pre-trained network with another fully connected layers, in which the output size represents the class of plants. Secondly, we use transfer experience and fine-tuned pre-trained DCNN for experiments using flower and fruit images. Finally, we evaluate the proposed network on two available botanical datasets: the Oxford flowers dataset with 102 classes and the HNPlant flowers and fruits dataset with 20 classes, and determine the optimal values of the associated hyperparameters to improve the overall performance. Experiment results demonstrate that the highest classification accuracies exhibited by the proposed model on the Oxford-102 and HNPlant-20 datasets are 92.4% and 95.0%, respectively, thus establishing their effectiveness and superiority.

\n
\n
\n

Plant diseases and pests detection based on deep learning: a review
\n Plant diseases and pests are important factors determining the yield and quality of plants. Plant diseases and pests identification can be carried out by means of digital image processing. In recent years, deep learning has made breakthroughs in the field of digital image processing, far superior to traditional methods. How to use deep learning technology to study plant diseases and pests identification has become a research issue of great concern to researchers. This review provides a definition of plant diseases and pests detection problem, puts forward a comparison with traditional plant diseases and pests detection methods. According to the difference of network structure, this study outlines the research on plant diseases and pests detection based on deep learning in recent years from three aspects of classification network, detection network and segmentation network, and the advantages and disadvantages of each method are summarized. Common datasets are introduced, and the performance of existing studies is compared. On this basis, this study discusses possible challenges in practical applications of plant diseases and pests detection based on deep learning. In addition, possible solutions and research ideas are proposed for the challenges, and several suggestions are given. Finally, this study gives the analysis and prospect of the future trend of plant diseases and pests detection based on deep learning.

\n
\n
\n

Extended application of deep learning combined with 2DCOS: Study on origin identification in the medicinal plant of Paris polyphylla var. yunnanensis.
\n INTRODUCTION
\n Medicinal plants are very important to human health, and ensuring their quality and rapid evaluation are the current research concerns. Deep learning has a strong ability in recognition. This study extended it to the identification of medicinal plants from the perspective of spectrum.
\n OBJECTIVE
\n In order to realise the rapid identification and provide a reference for the selection of high-quality resources of medicinal plants, a combination of deep learning and two-dimensional correlation spectroscopy (2DCOS) was proposed.
\n METHODS
\n For the first time, Fourier transform mid-infrared (FT-MIR) and near-infrared (NIR) spectroscopy 2DCOS images combined with residual neural network (ResNet) was used for the origin identification of Paris polyphylla var. yunnanensis. In total 1593 samples were collected and 12821 2DCOS images were drawn. The climate of different origins was briefly analysed.
\n RESULTS
\n The xishuangbanna, puer, lincang, honghe and wenshan are the five regions with more ecological advantages. The synchronous 2DCOS models of FT-MIR and NIR could realise origin identification with the accuracy of 100%. The synchronous images were suitable for the identification of medicinal plants with complex systems. The full band, feature band and different contour models had no big difference in distinguishing ability, so they were not the key factors affecting the discrimination results.
\n CONCLUSION
\n The ResNet models established were stable, reliable, and robust, which not only solved the problem of origin identification, expanded the application field of deep learning, but also provided practical reference for the related research of other medicinal plants.

\n
\n
\n

Plant Disease Identification Using Deep Learning: A Systematic Review
\n Agriculture contributes majorly in the Indian economy being the most important aspect of it. Often the plants suffer from many diseases which may be dependent on climatic conditions and pests which further degrades the quality of the crop. Early and accurate detection of disease in plants is very crucial step in depicting the overall yield of the crop, as this can increase the yield and productivity of the crop by a great margin. In the current climatic conditions, to obtain the superior quality crop is getting difficult day by day as the plants suffer from different diseases. To solve the problem of early and accurate detection image processing has come up with various techniques to find best and suitable ways. This paper presents a review to examine the power of these techniques in detection for plant diseases and add in the agriculture advancement. This survey enfolds larger scope of deep learning in the future research while detecting plant diseases along with improvised performance and accuracy. This paper also states certain challenges which still exist in the detection of diseases in plants and opens some areas of research for the researchers. Various problems in the field of dataset collection have been addressed. Some possible solutions also have been suggested which can help the accuracy to increase for a model.

\n
\n
\n

Plant Diseases Identification through a Discount Momentum Optimizer in Deep Learning
\n Deep learning proves its promising results in various domains. The automatic identification of plant diseases with deep convolutional neural networks attracts a lot of attention at present. This article extends stochastic gradient descent momentum optimizer and presents a discount momentum (DM) deep learning optimizer for plant diseases identification. To examine the recognition and generalization capability of the DM optimizer, we discuss the hyper-parameter tuning and convolutional neural networks models across the plantvillage dataset. We further conduct comparison experiments on popular non-adaptive learning rate methods. The proposed approach achieves an average validation accuracy of no less than 97% for plant diseases prediction on several state-of-the-art deep learning models and holds a low sensitivity to hyper-parameter settings. Experimental results demonstrate that the DM method can bring a higher identification performance, while still maintaining a competitive performance over other non-adaptive learning rate methods in terms of both training speed and generalization.

\n
\n
\n

A Study of Adversarial Attacks and Detection on Deep Learning-Based Plant Disease Identification
\n TL;DR: We found that successful adversarial attacks on deep learning (DL)-based plant disease identification systems could result in a significant delay of treatments and huge economic losses. Adversarial attacks with a small number of perturbations can dramatically degrade the performance of DNN models for plant disease identification. Adversarial attacks can be effectively defended by using adversarial sample detection with an appropriate choice of features.
\n Transfer learning using pre-trained deep neural networks (DNNs) has been widely used for plant disease identification recently. However, pre-trained DNNs are susceptible to adversarial attacks which generate adversarial samples causing DNN models to make wrong predictions. Successful adversarial attacks on deep learning (DL)-based plant disease identification systems could result in a significant delay of treatments and huge economic losses. This paper is the first attempt to study adversarial attacks and detection on DL-based plant disease identification. Our results show that adversarial attacks with a small number of perturbations can dramatically degrade the performance of DNN models for plant disease identification. We also find that adversarial attacks can be effectively defended by using adversarial sample detection with an appropriate choice of features. Our work will serve as a basis for developing more robust DNN models for plant disease identification and guiding the defense against adversarial attacks.

\n
\n
\n

Black Measles Disease Identification in Grape Plant (Vitis vinifera) Using Deep Learning
\n The most common diseases found in plants are the fungi infections/diseases. One of the common fungal diseases is Esca (Black Measles) which is found in the Grape Plants and can be easily identified as brown streaking lesions on any part of the leaf. The affected leaves can dry off completely and fall off from the plant prematurely which eventually results in death of the plant. In this work, an improved technique based on Deep Learning algorithm for identifying Esca Black measles in GrapeVines is proposed. The proposed method yields better performance and accuracy in detecting the disease, than past Machine Learning based approaches. Grape Plant dataset from PlantVillage Database is used for the work. The dataset contains total 1807 images (healthy and diseased). ResNet 50 architecture of Deep Neaural Network in combination with Transfer Learning and Fine Tuning was used to compute the results. The proposed system provides an accuracy of more than 97% and performed better than the existing approaches which are based on feature extraction methods.

\n
\n
\n

Identification of Plant Stomata Based on YOLO v5 Deep Learning Model
\n Stomata is an important structure in all terrestrial plants and is very vital in controlling plant photosynthesis and transpiration flow. Precise detection of plant stomata is the basis for studying stomata characteristics. Traditional detection methods are mostly manual operations, which is a tedious and inefficient process. Manually extracting features requires high image quality. Choosing appropriate features depends on certain prior knowledge, especially for the object with large morphological changes such as plant stomata. With the widespread use of deep learning technology, efficient solutions to this task have become possible. This article combines the characteristics of the corn leaf stomatal data sets to improve the latest object detection model YOLO v5)You Only Look Once(. By introducing the attention mechanism, that is, adding the SE module to the backbone network, the precision and recall of stoma detection are improved. At the same time, The loss function has been improved from to for avoiding some problems that may occur when selecting the best prediction box. Experimental results show that the precision and recall rates of the improved model on the corn leaf stomata data sets have reached 94.8% and 98.7% respectively, lay the foundation for the measurement of stomatal parameters. In addition, this paper also can help agriculturists and botanists to build their own data sets for stomatal research by explaining the methods of acquiring, pre-processing, and annotating data sets.

\n
\n
\n

Detection of plant leaf diseases using recent progress in Deep Learning-Based identification techniques
\n Mostly economy profoundly depends on farming efficiency. The farming crops are commonly affected by the disease. Since the economy depends on agriculture, this is one of the core reasons that infection identification in plants assumes a significant job in the horticulture field. On the off chance that legitimate consideration isn't taken here, at that point, it causes natural consequences for plants and because of which particular item quality, amount, or efficiency are influence. Crop misfortune because of ailments considerably influences the economy and undermines food accessibility. Quick and precise plant ailment location is essential to expanding farming efficiency in a supportable manner. In any case, plant location by human specialists is costly, tedious, and sometimes unrealistic. To counter these difficulties, Plant pathologists want an exact and dependable plant sickness conclusion framework. The on-going utilization of deep learning procedure with image processing methods for plant sickness acknowledgment has become a hot examination subject to give programmed analysis. This research provides a productive plant illness distinguishing proof technique dependent on pre-prepared deep learning models, such as AlexNet and GoogleNet designs. We trust that this work will be a significant asset for analysts in the area of ailment acknowledgment utilizing image handling strategies with deep learning architectures.

\n
\n
\n

Identification of Rice Plant Diseases Using Image Processing, Machine Learning & Deep Learning: A Review
\n TL;DR: This article discusses different methods for detecting rice plant diseases, with a focus on deep learning methods. It is found that deep learning methods are more promising than other methods for this task.
\n Agriculture is the primary source of livelihood for about more than 50% of the Indian population and rice is one of the major food grains of India. It is observed that rice plant diseases are the major contributors to reduce the production & quality of food. Identification of such diseases may improve the production quality. This paper gives an idea about different methods such as image processing, machine learning & deep learning which are used to detect deadly diseases in rice plants. Much research has been done to automate the rice plant disease detection process using images of the leaf. This manuscript has compared different rice plant disease detection methods and it is found that deep learning methods are more promising than other two methods.

\n
\n
\n

Plant Disease Identification on Real-World Data using Deep Learning: A Comparative Study
\n Correct and timely identification of disease in plants, especially for a country like India, where agriculture continues to serve as a cornerstone of its economy, is an indispensable tool demanding our solicitousness. Recently, researchers have started to employ autonomous real-time systems involving deep learning techniques for this purpose. However, the wide va-riety of heterogeneous diseases affecting crop yield continues to prove itself a mammoth task for farmers and stymies the researchers. The majority of the current state-of-the-art models utilize datasets like Plant Village, consisting of leaf images taken in a controlled lab environment, which do not serve as accurate representative data of the real-world scenario. Moreover, the effectiveness of state-of-the-art models like EfficientNetLite, which provided notable improvements in accuracy for similar deep learning applications, remains untested on plant disease datasets. Hence, an exhaustive study on the performance of various state-of-the-art models with varied training conditions on real-time datasets like PlantDoc is imperative to further this area of research. In this paper, we have explored state-of-the-art CNNs like InceptionResNet, EfficientNetLite_0, and VGG-19, under various parameter settings, image augmentations techniques, and loss functions on the real-time PlantDoc dataset. We have evaluated and presented a comparative analysis of the exhaustive combinations and performances gauged by accuracy, top-5 accuracy, F1 scores, and other inferences drawn from training and testing all these networks on 27 different classes of crop diseases. We infer that EfficientNetLite proved to be a most effective architecture, especially given its relatively smaller size. EfficientNetLite coupled with the focal loss function and Albumentaions augmentation library yielded the best results with an accuracy of 70.71%, mean F1-score as 0.7, and 95.57% top-5 accuracy, on a test set congruent with real-world relevance.

\n
\n
\n

Ensemble Deep Learning Models for Fine-grained Plant Species Identification
\n TL;DR: This study investigates the development and application of ensemble deep learning models for fine-grained plant species identification. Six popular deep learning models (InceptionResNetV2, ResNet50, Xception, InceptionV3, MobileNetV2, and GoogleNet) were trained on both datasets and heterogeneous ensembles with various ensemble techniques (mean, weighted mean, voting, and stacked generalization) were performed. The validation and testing accuracy results for individual models were compared with the output generated by the ensemble methods.
\n Automated plant species identification for the datasets (images) collected from the natural environment is a challenging task. This study investigates the development and application of ensemble deep learning models for fine-grained plant species identification. Two different types of plant species datasets have been used in this study. The first dataset (UBD_45) consists of 45 medicinal plant species from the natural environment with the imbalanced distribution of classes and the second dataset (VP_200) has 200 medicinal plant species with balanced classes from the natural environment. Six popular deep learning models (InceptionResNetV2, ResNet50, Xception, InceptionV3, MobileNetV2, and GoogleNet) were trained on both datasets and heterogeneous ensembles with various ensemble techniques (mean, weighted mean, voting, and stacked generalization) were performed. The validation and testing accuracy results for individual models were compared with the output generated by the ensemble methods. The highest testing accuracies for base models were found 96.7% and 91.2% for UBD_45 and VP_200 datasets, respectively. Mean, weighted mean, and stacking ensembles showed better performance for both datasets. The stacking ensemble improved the classification accuracy by around 1.8% for the UBD_45 dataset while for VP_200 a significant improvement of around 4.23% was noticed using a weighted mean ensemble.

\n
\n
\n

A Deep Learning Approach for Plant Material Disease Identification
\n Plant Material Disease Identification is essential for the food safety. To increase the crop production for the growing population of the world, the proper treatment is required on proper time to save the plant. Therefore, disease diagnosing on time is very important. This paper uses a deep learning convolutional neural network model to identify the plant disease. The pre-existing deep learning model Alexnet has been employed for plant disease identification in which an external feature of segmented plant material (leaves) is passed to the deepest fully connected layer. This combination of extracted feature by Alexnet and external feature of segmented plant material helps in plant disease identification. Experimental analysis has been done on a standard dataset Plant Village which has total 54,306 leaf images of 15 distinct plants having 38 diseases. The presented CNN approach worked well and outperformed to the existing approach.

\n
\n
\n

Plant Disease Identification Based on Deep Learning Algorithm in Smart Farming
\n The identification of plant disease is the premise of the prevention of plant disease efficiently and precisely in the complex environment. With the rapid development of the smart farming, the identification of plant disease becomes digitalized and data-driven, enabling advanced decision support, smart analyses, and planning. This paper proposes a mathematical model of plant disease detection and recognition based on deep learning, which improves accuracy, generality, and training efficiency. Firstly, the region proposal network (RPN) is utilized to recognize and localize the leaves in complex surroundings. Then, images segmented based on the results of RPN algorithm contain the feature of symptoms through Chan–Vese (CV) algorithm. Finally, the segmented leaves are input into the transfer learning model and trained by the dataset of diseased leaves under simple background. Furthermore, the model is examined with black rot, bacterial plaque, and rust diseases. The results show that the accuracy of the method is 83.57%, which is better than the traditional method, thus reducing the influence of disease on agricultural production and being favorable to sustainable development of agriculture. Therefore, the deep learning algorithm proposed in the paper is of great significance in intelligent agriculture, ecological protection, and agricultural production.

\n
\n
", "messages": [ { "id": 1770551, "postDate": "2022-04-28T10:59:35.113Z", "content": "

These are the papers that will change your life.
\nThese are the papers that will make you question everything you know.
\nThese are the papers that will make you see the world in a whole new way.
\nThese are the papers that will give you hints on improving your LB score ;)
\nHave fun!

\n
\n

Medicinal Plant Identification using Gabor Filters and Deep Learning Techniques: A Paper Review

\n
\n

TL;DR: Computer-aided identification of plants is a branch of machine learning that has become more recognized recently and proves itself as a vital tool in numerous sectors including pharmacological science, forestry and agriculture. This study reviewed plant species classification relying on leaf textural features using Gabor filters and revealed that Gabor filters perform better when combined with other feature extraction methods. Therefore, this study proposes using Log-Gabor filter in the field of plant identification to improve accuracy since they overcome the drawbacks of Gabor filters which are; the maximum bandwidth of a Gabor filter is limited to approximately one octave and Gabor filters are not optimal if one
\n Corresponding Author: Stephen Opoku Oppong ICT Education Department, University of Education, Winneba, Ghana Email: sooppong@uew.edu.gh Abstract: Computer-aided identification of plants is a branch of machine learning that has become more recognized recently and proves itself as a vital tool in numerous sectors including pharmacological science, forestry and agriculture. This has essentially generated a zeal in creating automated systems for the identification of diverse species of plants. This study reviewed plant species classification relying on leaf textural features using Gabor filters and revealed that Gabor filters perform better when combined with other feature extraction methods. Therefore, this study proposes using Log-Gabor filter in the field of plant identification to improve accuracy since they overcome the drawbacks of Gabor filters which are; the maximum bandwidth of a Gabor filter is limited to approximately one octave and Gabor filters are not optimal if one is seeking broad spectral information with maximal spatial localization.

\n
\n
\n

Holistic Based Plant Identification Using Deep Learning
\n TL;DR: This paper introduces an automated plant classification framework for recognizing plants dependent on their holistic shape. The system uses image preprocessing, deep learning techniques, and ResNet18 to identify plants through the camera; it will give their names in Urdu, English, and the plant's scientific name as an output. Moreover, the system can describe the plant along with its audio. The methodology was evaluated using two datasets. The first dataset, GRANDYMU, consists of 70 different plant species containing 3500 images, and the second is the famous SWEDISH leaf dataset. We trained different models based on Resnet18
\n Plants are the most important forms of life on earth. They are the fundamental asset for human prosperity, furnishing us with oxygen and food. Humans have widely utilized plants in medication, food formation, and the cosmetic business worldwide in their daily life activities. Identification of plants is extremely a challenging task due to the recursive nature of the shapes of plants. Moreover, identification has been done on a few categories and is limited to veins or leaf patterns. This paper introduced an automated plant classification framework for recognizing plants dependent on their holistic shape. The system uses image preprocessing, deep learning techniques, and ResNet18 to identify plants through the camera; it will give their names in Urdu, English, and the plant's scientific name as an output. Moreover, the system can describe the plant along with its audio. The methodology was evaluated using two datasets. The first dataset, GRANDYMU, consists of 70 different plant species containing 3500 images, and the second is the famous SWEDISH leaf dataset. We trained different models based on Resnet18, and the maximum accuracy that we achieved was 99%. Also, we got an accuracy of 99.9% for the latter dataset.

\n
\n
\n

Experimental evaluation of Data Augmentation heuristics for plant identification systems based on Deep Learning
\n Data augmentation (DA) allows increasing datasets for training machine learning models that demands large amounts of data. In real-world applications in which data may not be abundant enough and data acquisition is not easy, DA enables increasing diversity and introducing model generalization. In this work we evaluate several DA techniques and combining approaches to extend image datasets used to train plant species recognition models. We experimentally validated Deep Convolutional Neural Networks (DCNN) with several datasets obtained from common augmentation techniques and combinations. The results allowed the identification of the Translate + Crop augmentation policy as the most effective within the scope of evaluation.

\n
\n
\n

Deep Learning for Plant Identification in Natural Environment
\n TL;DR: We collected a dataset of 10,000 images of 100 ornamental plant species in Beijing Forestry University campus using mobile phones. A 26-layer deep learning model consisting of 8 residual building blocks was designed for large-scale plant classification in natural environment. The proposed model achieved a recognition rate of 91.78% on the BJFU100 dataset, demonstrating that deep learning is a promising technology for smart forestry.
\n Plant image identification has become an interdisciplinary focus in both botanical taxonomy and computer vision. The first plant image dataset collected by mobile phone in natural scene is presented, which contains 10,000 images of 100 ornamental plant species in Beijing Forestry University campus. A 26-layer deep learning model consisting of 8 residual building blocks is designed for large-scale plant classification in natural environment. The proposed model achieves a recognition rate of 91.78% on the BJFU100 dataset, demonstrating that deep learning is a promising technology for smart forestry.

\n
\n
\n

Plant Identification Based on Noisy Web Data: the Amazing Performance of Deep Learning (LifeCLEF 2017)
\n The 2017 fith edition of the LifeCLEF plant identification challenge is an important milestone towards automated plant identification systems working at the scale of continental floras with 10.000 plant species living mainly in Europe and North America illustrated by a total of 1.1M images. Nowadays, such ambitious systems are enabled thanks to the conjunction of the dazzling recent progress in image classification with deep learning and several outstanding international initiatives, such as the Encyclopedia of Life (EOL), aggregating the visual knowledge on plant species coming from the main national botany institutes. However, despite all these efforts the majority of the plant species still remain without pictures or are poorly illustrated. Outside the institutional channels, a much larger number of plant pictures are available and spread on the web through botanist blogs, plant lovers web-pages, image hosting websites and on-line plant retailers. The LifeCLEF 2017 plant challenge presented in this paper aimed at evaluating to what extent a large noisy training dataset collected through the web and containing a lot of labelling errors can compete with a smaller but trusted training dataset checked by experts. To fairly compare both training strategies, the test dataset was created from a third data source, i.e. the Pl@ntNet mobile application that collects millions of plant image queries all over the world. This paper presents more precisely the resources and assessments of the challenge, summarizes the approaches and systems employed by the participating research groups, and provides an analysis of the main outcomes. (Resume d'auteur)

\n
\n
\n

Deep Learning with Taxonomic Loss for Plant Identification
\n TL;DR: Train a neural network on a plant identification dataset using a taxonomic loss function to improve performance.
\n Plant identification is a fine-grained classification task which aims to identify the family, genus, and species according to plant appearance features. Inspired by the hierarchical structure of taxonomic tree, the taxonomic loss was proposed, which could encode the hierarchical relationships among multilevel labels into the deep learning objective function by simple group and sum operation. By training various neural networks on PlantCLEF 2015 and PlantCLEF 2017 datasets, the experimental results demonstrated that the proposed loss function was easy to implement and outperformed the most commonly adopted cross-entropy loss. Eight neural networks were trained, respectively, by two different loss functions on PlantCLEF 2015 dataset, and the models trained by taxonomic loss led to significant performance improvements. On PlantCLEF 2017 dataset with 10,000 species, the SENet-154 model trained by taxonomic loss achieved the accuracies of 84.07%, 79.97%, and 73.61% at family, genus and species levels, which improved those of model trained by cross-entropy loss by 2.23%, 1.34%, and 1.08%, respectively. The taxonomic loss could further facilitate the fine-grained classification task with hierarchical labels.

\n
\n
\n

A Light Weight Deep Learning Model for Real World Plant Identification
\n Automatic identification and classification of different plant leaf species have become a common trend among researchers and scientists. To obtain a result with better precision, they use various methods and techniques of deep learning to build a model. Convolutional neural networks are becoming the most common method used by scientists to classify plant leaves. However, the classification of plant leaves can be challenging with more rare species and complicated backgrounds, for which researchers build several models to achieve high-level accuracy. In the present study for the classification of leaves, we have created a model for plant leaf classification based on a dataset we collected. We've used the Resnet-50 model, a well-known CNN architecture, which provided an efficient method to organize and analyze a deep classification to reduce the complexity so that there will be fewer parameters for training and low time consumption as well. Using Resnet-50, we intended to develop a significant result in our classification model. The convolutional neural network is famous for its influential abilities in feature extraction and classification. And Resnet-50 being a residual network enabled us to train deep networks in our model. The average training accuracy reached 98.3%, while the average testing accuracy reached 92.5%. The key contribution of this study is effective accuracy as well as we have trained the model on our own prepared dataset that we have prepared from real world environment. Data Availability: https://drive.google.com/file/d/1bD7B257l-6wqUCQHBWhle95xyrotUbwO/view?usp=sharing

\n
\n
\n

Fine-Grained Plant Identification using wide and deep learning model 1
\n TL;DR: We propose a model to address the fine-grained plant image classification task by using the wide and deep learning framework which combines a linear model and a deep learning model. Proposed method sums the result of the wide and deep learning model using a logistic function so that discrete features can be considered simultaneously with continuous image content.
\n In recent years, with the evolution of deep learning technology, the performance of plant image recognition has improved remarkably. In this paper, we propose a model to address the fine-grained plant image classification task by using the wide and deep learning framework which combines a linear model and a deep learning model. Proposed method sums the result of the wide and deep learning model using a logistic function so that discrete features can be considered simultaneously with continuous image content. Our works used metadata such as the date of flowering and locational information for the wide model. Our experiment shows that the proposed method gives better performance than a baseline method.

\n
\n
\n

Plant Identification: Experts vs. Machines in the Era of Deep Learning - Deep Learning Techniques Challenge Flora Experts
\n TL;DR: Nine deep learning models were evaluated with regard to nine French botanists. The performance of the deep learning models was found to be close to the human expertise.
\n Automated identification of plants and animals have improved considerably in the last few years, in particular thanks to the recent advances in deep learning. The next big question is how far such automated systems are from the human expertise. Indeed, even the best experts are sometimes confused and/or disagree between each others when validating visual or audio observations of living organism. A picture or a sound actually contains only a partial information that is usually not sufficient to determine the right species with certainty. Quantifying this uncertainty and comparing it to the performance of automated systems is of high interest for both computer scientists and expert naturalists. This chapter reports an experimental study following this idea in the plant domain. In total, nine deep-learning systems implemented by three different research teams were evaluated with regard to nine expert botanists of the French flora. Therefore, we created a small set of plant observations that were identified in the field and revised by experts in order to have a near-perfect golden standard. The main outcome of this work is that the performance of state-of-the-art deep learning models is now close to the most advanced human expertise. This shows that automated plant identification systems are now mature enough for several routine tasks, and can offer very promising tools for autonomous ecological surveillance systems.

\n
\n
\n

Observation on Therapeutic Plant Identification based on Deep Learning Technique
\n TL;DR: There are many different methods for leaf classification, but the most successful ones rely on deep learning algorithms. These algorithms can learn to identify patterns in data that are too complex for humans to discern. This makes them very effective at identifying different types of leaves.
\n Plants have been used for medicinal purposes long before recorded history. It plays a major role in medicines, food, perfumes and cosmetics industries. By knowing the herbal plants and its usage it can be used for above applications. In this digital era, people don’t have adequate knowledge to identify various herbal plants which are used by our ancestors for long time. Presently, the identification of herbal plants is purely based on the human perception or knowledge. There may be probability of human error occurring. In order to have an efficient herb species classification, there must be a complete model which should be automatic and convenient recognition system. This paper is reviewing the different leaf classification methodologies based on deep learning algorithms. The main aim of this research paper is to conclude the advanced technique for the leaf identification.

\n
\n
\n

Deep learning for plant identification: how the web can compete with human experts
\n Automated identification of plants and animals has improved considerably in the last few years, in particular thanks to the recent advances in deep learning. In order to evaluate the performance of automated plant identification technologies in a sustainable and repeatable way, a dedicated system-oriented benchmark was setup in 2011 in the context of ImageCLEF (Goëau et al. 2011). Each year, since that time, several research groups participated in this large collaborative evaluation by benchmarking their image-based plant identification systems. In 2014, the LifeCLEF research platform (Joly et al. 2014) was created in the continuity of this effort so as to enlarge the evaluated challenges by considering birds and fishes in addition to plants, and audio and video contents in addition to images. The 2017-th edition of the LifeCLEF plant identification challenge (Joly et al. 2017) is an important milestone towards automated plant identification systems working at the scale of continental floras with 10.000 plant species living mainly in Europe and North America ‡,§ | §,‡ ¶ # ¤ © Goëau H et al. This is an open access article distributed under the terms of the Creative Commons Attribution License (CC BY 4.0), which permits unrestricted use, distribution, and reproduction in any medium, provided the original author and source are credited. illustrated by a total of 1.1M images. Nowadays, such ambitious systems are enabled thanks to the conjunction of the dazzling recent progress in image classification with deep learning and several outstanding international initiatives, aggregating the visual knowledge on plant species coming from the main national botanical institutes. The PlantCLEF plant challenge that we propose to present at this workshop aimed at evaluating to what extent a large noisy training dataset collected through the web (then containing a lot of labelling errors) can compete with a smaller but trusted training dataset checked by experts. To fairly compare both training strategies, the test dataset was created from a third data source, the Pl@ntNet (Joly et al. 2015) mobile application that collects millions of plant image queries all over the world. Due to the good results obtained at the 2017-th edition of the LifeCLEF plant identification challenge, the next big question is how far such automated systems are from the human expertise. Indeed, even the best experts are sometimes confused and/or disagree with each other when validating images of living organism. A multimedia data actually contains only partial information that is usually not sufficient to determine the right species with certainty. Quantifying this uncertainty and comparing it to the performance of automated systems is of high interest for both computer scientists and expert naturalists. This work reports an experimental study following this idea in the plant domain. In total, 9 deeplearning systems implemented by 3 different research teams were evaluated with regard to 9 expert botanists of the French flora. The main outcome of this work is that the performance of state-of-the-art deep learning models is now close to the most advanced human expertise. This shows that automated plant identification systems are now mature enough for several routine tasks, and can offer very promising tools for autonomous ecological surveillance systems.

\n
\n
\n

Plant Identification with Deep Learning Ensembles
\n This work describes the plant identification system that we submitted to the ExpertLifeCLEF plant identification campaign in 2018. We fine-tuned two pre-trained deep learning architectures (SeNet and DensNetwork) using images shared by the CLEF organizers in 2017. Our main runs are 4 ensembles obtained with different weighted combinations of the 4 deep learning architectures. The fifth ensemble is based on deep learning features but uses Error Correcting Output Codes (ECOC) as the ensemble. Our best system has achieved a classification accuracy of 74.4%, while the best system obtained 86.7% accuracy, on the whole of the official test data. This system ranked 4th place among all the teams, but matched the accuracy of one of the human experts.

\n
\n
\n

A Combination of Deep Learning and Hand-Designed Feature for Plant Identification Based on Leaf and Flower Images
\n TL;DR: This paper proposes a combination of deep learning and hand-designed feature for plant identification based on leaf and flower images. The contributions of this paper are two-fold. First, for each organ image, we have performed a comparative evaluation of deep learning and hand-designed feature for plant identification. Two approaches for deep learning and hand-designed feature that are convolutional neuron network (CNN) and kernel descriptor (KDES) are chosen in our experiments. Second, based on the results of the first contribution, we propose a method for plant identification by late fusing the identification results of leaf and flower. Experimental results on ImageCle
\n This paper proposes a combination of deep learning and hand-designed feature for plant identification based on leaf and flower images. The contributions of this paper are two-fold. First, for each organ image, we have performed a comparative evaluation of deep learning and hand-designed feature for plant identification. Two approaches for deep learning and hand-designed feature that are convolutional neuron network (CNN) and kernel descriptor (KDES) are chosen in our experiments. Second, based on the results of the first contribution, we propose a method for plant identification by late fusing the identification results of leaf and flower. Experimental results on ImageClef 2015 dataset show that hand designed feature outperforms deep learning for well-constrained cases (leaf captured on simple background). However, deep learning shows its robustness in natural situations. Moreover, the combination of leaf and flower images improves significantly the identification when comparing leaf-based plant identification.

\n
\n
\n

Transfer learning with deep convolutional neural network for automated plant identification in hainan island
\n The Hainan Island has a generally high biological diversity with a wide variety of plant species, some of which are listed as endemic to the island. It is time-consuming and difficult, even for the botanist experts to determine the name of species based on observations. Automated plant identification enables experts to process significantly greater numbers of plants with higher efficiencies in shorter periods of time. However, plant recognition is a kind of fine-grained visual recognition problem, which is relatively harder than conventional image recognition. In this paper, we employ a Deep Convolutional Neural Network (DCNN) trained on the ImageNet database, which contains millions of images, and then transfer the learning information for automated plant identification based on flower and fruit images. First, we modify the last three layers of the pre-trained network in order to adapt ResNet-50 model to our classification task, and replace the fully connected layer in the original pre-trained network with another fully connected layers, in which the output size represents the class of plants. Secondly, we use transfer experience and fine-tuned pre-trained DCNN for experiments using flower and fruit images. Finally, we evaluate the proposed network on two available botanical datasets: the Oxford flowers dataset with 102 classes and the HNPlant flowers and fruits dataset with 20 classes, and determine the optimal values of the associated hyperparameters to improve the overall performance. Experiment results demonstrate that the highest classification accuracies exhibited by the proposed model on the Oxford-102 and HNPlant-20 datasets are 92.4% and 95.0%, respectively, thus establishing their effectiveness and superiority.

\n
\n
\n

Plant diseases and pests detection based on deep learning: a review
\n Plant diseases and pests are important factors determining the yield and quality of plants. Plant diseases and pests identification can be carried out by means of digital image processing. In recent years, deep learning has made breakthroughs in the field of digital image processing, far superior to traditional methods. How to use deep learning technology to study plant diseases and pests identification has become a research issue of great concern to researchers. This review provides a definition of plant diseases and pests detection problem, puts forward a comparison with traditional plant diseases and pests detection methods. According to the difference of network structure, this study outlines the research on plant diseases and pests detection based on deep learning in recent years from three aspects of classification network, detection network and segmentation network, and the advantages and disadvantages of each method are summarized. Common datasets are introduced, and the performance of existing studies is compared. On this basis, this study discusses possible challenges in practical applications of plant diseases and pests detection based on deep learning. In addition, possible solutions and research ideas are proposed for the challenges, and several suggestions are given. Finally, this study gives the analysis and prospect of the future trend of plant diseases and pests detection based on deep learning.

\n
\n
\n

Extended application of deep learning combined with 2DCOS: Study on origin identification in the medicinal plant of Paris polyphylla var. yunnanensis.
\n INTRODUCTION
\n Medicinal plants are very important to human health, and ensuring their quality and rapid evaluation are the current research concerns. Deep learning has a strong ability in recognition. This study extended it to the identification of medicinal plants from the perspective of spectrum.
\n OBJECTIVE
\n In order to realise the rapid identification and provide a reference for the selection of high-quality resources of medicinal plants, a combination of deep learning and two-dimensional correlation spectroscopy (2DCOS) was proposed.
\n METHODS
\n For the first time, Fourier transform mid-infrared (FT-MIR) and near-infrared (NIR) spectroscopy 2DCOS images combined with residual neural network (ResNet) was used for the origin identification of Paris polyphylla var. yunnanensis. In total 1593 samples were collected and 12821 2DCOS images were drawn. The climate of different origins was briefly analysed.
\n RESULTS
\n The xishuangbanna, puer, lincang, honghe and wenshan are the five regions with more ecological advantages. The synchronous 2DCOS models of FT-MIR and NIR could realise origin identification with the accuracy of 100%. The synchronous images were suitable for the identification of medicinal plants with complex systems. The full band, feature band and different contour models had no big difference in distinguishing ability, so they were not the key factors affecting the discrimination results.
\n CONCLUSION
\n The ResNet models established were stable, reliable, and robust, which not only solved the problem of origin identification, expanded the application field of deep learning, but also provided practical reference for the related research of other medicinal plants.

\n
\n
\n

Plant Disease Identification Using Deep Learning: A Systematic Review
\n Agriculture contributes majorly in the Indian economy being the most important aspect of it. Often the plants suffer from many diseases which may be dependent on climatic conditions and pests which further degrades the quality of the crop. Early and accurate detection of disease in plants is very crucial step in depicting the overall yield of the crop, as this can increase the yield and productivity of the crop by a great margin. In the current climatic conditions, to obtain the superior quality crop is getting difficult day by day as the plants suffer from different diseases. To solve the problem of early and accurate detection image processing has come up with various techniques to find best and suitable ways. This paper presents a review to examine the power of these techniques in detection for plant diseases and add in the agriculture advancement. This survey enfolds larger scope of deep learning in the future research while detecting plant diseases along with improvised performance and accuracy. This paper also states certain challenges which still exist in the detection of diseases in plants and opens some areas of research for the researchers. Various problems in the field of dataset collection have been addressed. Some possible solutions also have been suggested which can help the accuracy to increase for a model.

\n
\n
\n

Plant Diseases Identification through a Discount Momentum Optimizer in Deep Learning
\n Deep learning proves its promising results in various domains. The automatic identification of plant diseases with deep convolutional neural networks attracts a lot of attention at present. This article extends stochastic gradient descent momentum optimizer and presents a discount momentum (DM) deep learning optimizer for plant diseases identification. To examine the recognition and generalization capability of the DM optimizer, we discuss the hyper-parameter tuning and convolutional neural networks models across the plantvillage dataset. We further conduct comparison experiments on popular non-adaptive learning rate methods. The proposed approach achieves an average validation accuracy of no less than 97% for plant diseases prediction on several state-of-the-art deep learning models and holds a low sensitivity to hyper-parameter settings. Experimental results demonstrate that the DM method can bring a higher identification performance, while still maintaining a competitive performance over other non-adaptive learning rate methods in terms of both training speed and generalization.

\n
\n
\n

A Study of Adversarial Attacks and Detection on Deep Learning-Based Plant Disease Identification
\n TL;DR: We found that successful adversarial attacks on deep learning (DL)-based plant disease identification systems could result in a significant delay of treatments and huge economic losses. Adversarial attacks with a small number of perturbations can dramatically degrade the performance of DNN models for plant disease identification. Adversarial attacks can be effectively defended by using adversarial sample detection with an appropriate choice of features.
\n Transfer learning using pre-trained deep neural networks (DNNs) has been widely used for plant disease identification recently. However, pre-trained DNNs are susceptible to adversarial attacks which generate adversarial samples causing DNN models to make wrong predictions. Successful adversarial attacks on deep learning (DL)-based plant disease identification systems could result in a significant delay of treatments and huge economic losses. This paper is the first attempt to study adversarial attacks and detection on DL-based plant disease identification. Our results show that adversarial attacks with a small number of perturbations can dramatically degrade the performance of DNN models for plant disease identification. We also find that adversarial attacks can be effectively defended by using adversarial sample detection with an appropriate choice of features. Our work will serve as a basis for developing more robust DNN models for plant disease identification and guiding the defense against adversarial attacks.

\n
\n
\n

Black Measles Disease Identification in Grape Plant (Vitis vinifera) Using Deep Learning
\n The most common diseases found in plants are the fungi infections/diseases. One of the common fungal diseases is Esca (Black Measles) which is found in the Grape Plants and can be easily identified as brown streaking lesions on any part of the leaf. The affected leaves can dry off completely and fall off from the plant prematurely which eventually results in death of the plant. In this work, an improved technique based on Deep Learning algorithm for identifying Esca Black measles in GrapeVines is proposed. The proposed method yields better performance and accuracy in detecting the disease, than past Machine Learning based approaches. Grape Plant dataset from PlantVillage Database is used for the work. The dataset contains total 1807 images (healthy and diseased). ResNet 50 architecture of Deep Neaural Network in combination with Transfer Learning and Fine Tuning was used to compute the results. The proposed system provides an accuracy of more than 97% and performed better than the existing approaches which are based on feature extraction methods.

\n
\n
\n

Identification of Plant Stomata Based on YOLO v5 Deep Learning Model
\n Stomata is an important structure in all terrestrial plants and is very vital in controlling plant photosynthesis and transpiration flow. Precise detection of plant stomata is the basis for studying stomata characteristics. Traditional detection methods are mostly manual operations, which is a tedious and inefficient process. Manually extracting features requires high image quality. Choosing appropriate features depends on certain prior knowledge, especially for the object with large morphological changes such as plant stomata. With the widespread use of deep learning technology, efficient solutions to this task have become possible. This article combines the characteristics of the corn leaf stomatal data sets to improve the latest object detection model YOLO v5)You Only Look Once(. By introducing the attention mechanism, that is, adding the SE module to the backbone network, the precision and recall of stoma detection are improved. At the same time, The loss function has been improved from to for avoiding some problems that may occur when selecting the best prediction box. Experimental results show that the precision and recall rates of the improved model on the corn leaf stomata data sets have reached 94.8% and 98.7% respectively, lay the foundation for the measurement of stomatal parameters. In addition, this paper also can help agriculturists and botanists to build their own data sets for stomatal research by explaining the methods of acquiring, pre-processing, and annotating data sets.

\n
\n
\n

Detection of plant leaf diseases using recent progress in Deep Learning-Based identification techniques
\n Mostly economy profoundly depends on farming efficiency. The farming crops are commonly affected by the disease. Since the economy depends on agriculture, this is one of the core reasons that infection identification in plants assumes a significant job in the horticulture field. On the off chance that legitimate consideration isn't taken here, at that point, it causes natural consequences for plants and because of which particular item quality, amount, or efficiency are influence. Crop misfortune because of ailments considerably influences the economy and undermines food accessibility. Quick and precise plant ailment location is essential to expanding farming efficiency in a supportable manner. In any case, plant location by human specialists is costly, tedious, and sometimes unrealistic. To counter these difficulties, Plant pathologists want an exact and dependable plant sickness conclusion framework. The on-going utilization of deep learning procedure with image processing methods for plant sickness acknowledgment has become a hot examination subject to give programmed analysis. This research provides a productive plant illness distinguishing proof technique dependent on pre-prepared deep learning models, such as AlexNet and GoogleNet designs. We trust that this work will be a significant asset for analysts in the area of ailment acknowledgment utilizing image handling strategies with deep learning architectures.

\n
\n
\n

Identification of Rice Plant Diseases Using Image Processing, Machine Learning & Deep Learning: A Review
\n TL;DR: This article discusses different methods for detecting rice plant diseases, with a focus on deep learning methods. It is found that deep learning methods are more promising than other methods for this task.
\n Agriculture is the primary source of livelihood for about more than 50% of the Indian population and rice is one of the major food grains of India. It is observed that rice plant diseases are the major contributors to reduce the production & quality of food. Identification of such diseases may improve the production quality. This paper gives an idea about different methods such as image processing, machine learning & deep learning which are used to detect deadly diseases in rice plants. Much research has been done to automate the rice plant disease detection process using images of the leaf. This manuscript has compared different rice plant disease detection methods and it is found that deep learning methods are more promising than other two methods.

\n
\n
\n

Plant Disease Identification on Real-World Data using Deep Learning: A Comparative Study
\n Correct and timely identification of disease in plants, especially for a country like India, where agriculture continues to serve as a cornerstone of its economy, is an indispensable tool demanding our solicitousness. Recently, researchers have started to employ autonomous real-time systems involving deep learning techniques for this purpose. However, the wide va-riety of heterogeneous diseases affecting crop yield continues to prove itself a mammoth task for farmers and stymies the researchers. The majority of the current state-of-the-art models utilize datasets like Plant Village, consisting of leaf images taken in a controlled lab environment, which do not serve as accurate representative data of the real-world scenario. Moreover, the effectiveness of state-of-the-art models like EfficientNetLite, which provided notable improvements in accuracy for similar deep learning applications, remains untested on plant disease datasets. Hence, an exhaustive study on the performance of various state-of-the-art models with varied training conditions on real-time datasets like PlantDoc is imperative to further this area of research. In this paper, we have explored state-of-the-art CNNs like InceptionResNet, EfficientNetLite_0, and VGG-19, under various parameter settings, image augmentations techniques, and loss functions on the real-time PlantDoc dataset. We have evaluated and presented a comparative analysis of the exhaustive combinations and performances gauged by accuracy, top-5 accuracy, F1 scores, and other inferences drawn from training and testing all these networks on 27 different classes of crop diseases. We infer that EfficientNetLite proved to be a most effective architecture, especially given its relatively smaller size. EfficientNetLite coupled with the focal loss function and Albumentaions augmentation library yielded the best results with an accuracy of 70.71%, mean F1-score as 0.7, and 95.57% top-5 accuracy, on a test set congruent with real-world relevance.

\n
\n
\n

Ensemble Deep Learning Models for Fine-grained Plant Species Identification
\n TL;DR: This study investigates the development and application of ensemble deep learning models for fine-grained plant species identification. Six popular deep learning models (InceptionResNetV2, ResNet50, Xception, InceptionV3, MobileNetV2, and GoogleNet) were trained on both datasets and heterogeneous ensembles with various ensemble techniques (mean, weighted mean, voting, and stacked generalization) were performed. The validation and testing accuracy results for individual models were compared with the output generated by the ensemble methods.
\n Automated plant species identification for the datasets (images) collected from the natural environment is a challenging task. This study investigates the development and application of ensemble deep learning models for fine-grained plant species identification. Two different types of plant species datasets have been used in this study. The first dataset (UBD_45) consists of 45 medicinal plant species from the natural environment with the imbalanced distribution of classes and the second dataset (VP_200) has 200 medicinal plant species with balanced classes from the natural environment. Six popular deep learning models (InceptionResNetV2, ResNet50, Xception, InceptionV3, MobileNetV2, and GoogleNet) were trained on both datasets and heterogeneous ensembles with various ensemble techniques (mean, weighted mean, voting, and stacked generalization) were performed. The validation and testing accuracy results for individual models were compared with the output generated by the ensemble methods. The highest testing accuracies for base models were found 96.7% and 91.2% for UBD_45 and VP_200 datasets, respectively. Mean, weighted mean, and stacking ensembles showed better performance for both datasets. The stacking ensemble improved the classification accuracy by around 1.8% for the UBD_45 dataset while for VP_200 a significant improvement of around 4.23% was noticed using a weighted mean ensemble.

\n
\n
\n

A Deep Learning Approach for Plant Material Disease Identification
\n Plant Material Disease Identification is essential for the food safety. To increase the crop production for the growing population of the world, the proper treatment is required on proper time to save the plant. Therefore, disease diagnosing on time is very important. This paper uses a deep learning convolutional neural network model to identify the plant disease. The pre-existing deep learning model Alexnet has been employed for plant disease identification in which an external feature of segmented plant material (leaves) is passed to the deepest fully connected layer. This combination of extracted feature by Alexnet and external feature of segmented plant material helps in plant disease identification. Experimental analysis has been done on a standard dataset Plant Village which has total 54,306 leaf images of 15 distinct plants having 38 diseases. The presented CNN approach worked well and outperformed to the existing approach.

\n
\n
\n

Plant Disease Identification Based on Deep Learning Algorithm in Smart Farming
\n The identification of plant disease is the premise of the prevention of plant disease efficiently and precisely in the complex environment. With the rapid development of the smart farming, the identification of plant disease becomes digitalized and data-driven, enabling advanced decision support, smart analyses, and planning. This paper proposes a mathematical model of plant disease detection and recognition based on deep learning, which improves accuracy, generality, and training efficiency. Firstly, the region proposal network (RPN) is utilized to recognize and localize the leaves in complex surroundings. Then, images segmented based on the results of RPN algorithm contain the feature of symptoms through Chan–Vese (CV) algorithm. Finally, the segmented leaves are input into the transfer learning model and trained by the dataset of diseased leaves under simple background. Furthermore, the model is examined with black rot, bacterial plaque, and rust diseases. The results show that the accuracy of the method is 83.57%, which is better than the traditional method, thus reducing the influence of disease on agricultural production and being favorable to sustainable development of agriculture. Therefore, the deep learning algorithm proposed in the paper is of great significance in intelligent agriculture, ecological protection, and agricultural production.

\n
\n
", "rawMarkdown": "\nThese are the papers that will change your life.\n\nThese are the papers that will make you question everything you know.\n\nThese are the papers that will make you see the world in a whole new way.\n\nThese are the papers that will give you hints on improving your LB score ;)\n\nHave fun!\n\n_____\n**[Medicinal Plant Identification using Gabor Filters and Deep Learning Techniques: A Paper Review](https://doi.org/10.3844/jcssp.2021.1210.1221)**\n\n\n> **TL;DR:** Computer-aided identification of plants is a branch of machine learning that has become more recognized recently and proves itself as a vital tool in numerous sectors including pharmacological science, forestry and agriculture. This study reviewed plant species classification relying on leaf textural features using Gabor filters and revealed that Gabor filters perform better when combined with other feature extraction methods. Therefore, this study proposes using Log-Gabor filter in the field of plant identification to improve accuracy since they overcome the drawbacks of Gabor filters which are; the maximum bandwidth of a Gabor filter is limited to approximately one octave and Gabor filters are not optimal if one\n\nCorresponding Author: Stephen Opoku Oppong ICT Education Department, University of Education, Winneba, Ghana Email: sooppong@uew.edu.gh Abstract: Computer-aided identification of plants is a branch of machine learning that has become more recognized recently and proves itself as a vital tool in numerous sectors including pharmacological science, forestry and agriculture. This has essentially generated a zeal in creating automated systems for the identification of diverse species of plants. This study reviewed plant species classification relying on leaf textural features using Gabor filters and revealed that Gabor filters perform better when combined with other feature extraction methods. Therefore, this study proposes using Log-Gabor filter in the field of plant identification to improve accuracy since they overcome the drawbacks of Gabor filters which are; the maximum bandwidth of a Gabor filter is limited to approximately one octave and Gabor filters are not optimal if one is seeking broad spectral information with maximal spatial localization.\n_____\n\n_____\n**[Holistic Based Plant Identification Using Deep Learning](https://doi.org/10.1109/icet54505.2021.9689804)**\n\n\n> **TL;DR:** This paper introduces an automated plant classification framework for recognizing plants dependent on their holistic shape. The system uses image preprocessing, deep learning techniques, and ResNet18 to identify plants through the camera; it will give their names in Urdu, English, and the plant's scientific name as an output. Moreover, the system can describe the plant along with its audio. The methodology was evaluated using two datasets. The first dataset, GRANDYMU, consists of 70 different plant species containing 3500 images, and the second is the famous SWEDISH leaf dataset. We trained different models based on Resnet18\n\nPlants are the most important forms of life on earth. They are the fundamental asset for human prosperity, furnishing us with oxygen and food. Humans have widely utilized plants in medication, food formation, and the cosmetic business worldwide in their daily life activities. Identification of plants is extremely a challenging task due to the recursive nature of the shapes of plants. Moreover, identification has been done on a few categories and is limited to veins or leaf patterns. This paper introduced an automated plant classification framework for recognizing plants dependent on their holistic shape. The system uses image preprocessing, deep learning techniques, and ResNet18 to identify plants through the camera; it will give their names in Urdu, English, and the plant's scientific name as an output. Moreover, the system can describe the plant along with its audio. The methodology was evaluated using two datasets. The first dataset, GRANDYMU, consists of 70 different plant species containing 3500 images, and the second is the famous SWEDISH leaf dataset. We trained different models based on Resnet18, and the maximum accuracy that we achieved was 99%. Also, we got an accuracy of 99.9% for the latter dataset.\n_____\n\n_____\n**[Experimental evaluation of Data Augmentation heuristics for plant identification systems based on Deep Learning](https://doi.org/10.5753/sbiagro.2021.18384)**\n\nData augmentation (DA) allows increasing datasets for training machine learning models that demands large amounts of data. In real-world applications in which data may not be abundant enough and data acquisition is not easy, DA enables increasing diversity and introducing model generalization. In this work we evaluate several DA techniques and combining approaches to extend image datasets used to train plant species recognition models. We experimentally validated Deep Convolutional Neural Networks (DCNN) with several datasets obtained from common augmentation techniques and combinations. The results allowed the identification of the Translate + Crop augmentation policy as the most effective within the scope of evaluation.\n_____\n\n_____\n**[Deep Learning for Plant Identification in Natural Environment](https://doi.org/10.1155/2017/7361042)**\n\n\n> **TL;DR:** We collected a dataset of 10,000 images of 100 ornamental plant species in Beijing Forestry University campus using mobile phones. A 26-layer deep learning model consisting of 8 residual building blocks was designed for large-scale plant classification in natural environment. The proposed model achieved a recognition rate of 91.78% on the BJFU100 dataset, demonstrating that deep learning is a promising technology for smart forestry.\n\nPlant image identification has become an interdisciplinary focus in both botanical taxonomy and computer vision. The first plant image dataset collected by mobile phone in natural scene is presented, which contains 10,000 images of 100 ornamental plant species in Beijing Forestry University campus. A 26-layer deep learning model consisting of 8 residual building blocks is designed for large-scale plant classification in natural environment. The proposed model achieves a recognition rate of 91.78% on the BJFU100 dataset, demonstrating that deep learning is a promising technology for smart forestry.\n_____\n\n_____\n**[Plant Identification Based on Noisy Web Data: the Amazing Performance of Deep Learning (LifeCLEF 2017)](https://scholar.google.com/scholar?q=Plant%20Identification%20Based%20on%20Noisy%20Web%20Data%3A%20the%20Amazing%20Performance%20of%20Deep%20Learning%20(LifeCLEF%202017))**\n\nThe 2017 fith edition of the LifeCLEF plant identification challenge is an important milestone towards automated plant identification systems working at the scale of continental floras with 10.000 plant species living mainly in Europe and North America illustrated by a total of 1.1M images. Nowadays, such ambitious systems are enabled thanks to the conjunction of the dazzling recent progress in image classification with deep learning and several outstanding international initiatives, such as the Encyclopedia of Life (EOL), aggregating the visual knowledge on plant species coming from the main national botany institutes. However, despite all these efforts the majority of the plant species still remain without pictures or are poorly illustrated. Outside the institutional channels, a much larger number of plant pictures are available and spread on the web through botanist blogs, plant lovers web-pages, image hosting websites and on-line plant retailers. The LifeCLEF 2017 plant challenge presented in this paper aimed at evaluating to what extent a large noisy training dataset collected through the web and containing a lot of labelling errors can compete with a smaller but trusted training dataset checked by experts. To fairly compare both training strategies, the test dataset was created from a third data source, i.e. the Pl@ntNet mobile application that collects millions of plant image queries all over the world. This paper presents more precisely the resources and assessments of the challenge, summarizes the approaches and systems employed by the participating research groups, and provides an analysis of the main outcomes. (Resume d'auteur)\n_____\n\n_____\n**[Deep Learning with Taxonomic Loss for Plant Identification](https://doi.org/10.1155/2019/2015017)**\n\n\n> **TL;DR:** Train a neural network on a plant identification dataset using a taxonomic loss function to improve performance.\n\nPlant identification is a fine-grained classification task which aims to identify the family, genus, and species according to plant appearance features. Inspired by the hierarchical structure of taxonomic tree, the taxonomic loss was proposed, which could encode the hierarchical relationships among multilevel labels into the deep learning objective function by simple group and sum operation. By training various neural networks on PlantCLEF 2015 and PlantCLEF 2017 datasets, the experimental results demonstrated that the proposed loss function was easy to implement and outperformed the most commonly adopted cross-entropy loss. Eight neural networks were trained, respectively, by two different loss functions on PlantCLEF 2015 dataset, and the models trained by taxonomic loss led to significant performance improvements. On PlantCLEF 2017 dataset with 10,000 species, the SENet-154 model trained by taxonomic loss achieved the accuracies of 84.07%, 79.97%, and 73.61% at family, genus and species levels, which improved those of model trained by cross-entropy loss by 2.23%, 1.34%, and 1.08%, respectively. The taxonomic loss could further facilitate the fine-grained classification task with hierarchical labels.\n_____\n\n_____\n**[A Light Weight Deep Learning Model for Real World Plant Identification](https://doi.org/10.1109/dchpc55044.2022.9731841)**\n\nAutomatic identification and classification of different plant leaf species have become a common trend among researchers and scientists. To obtain a result with better precision, they use various methods and techniques of deep learning to build a model. Convolutional neural networks are becoming the most common method used by scientists to classify plant leaves. However, the classification of plant leaves can be challenging with more rare species and complicated backgrounds, for which researchers build several models to achieve high-level accuracy. In the present study for the classification of leaves, we have created a model for plant leaf classification based on a dataset we collected. We've used the Resnet-50 model, a well-known CNN architecture, which provided an efficient method to organize and analyze a deep classification to reduce the complexity so that there will be fewer parameters for training and low time consumption as well. Using Resnet-50, we intended to develop a significant result in our classification model. The convolutional neural network is famous for its influential abilities in feature extraction and classification. And Resnet-50 being a residual network enabled us to train deep networks in our model. The average training accuracy reached 98.3%, while the average testing accuracy reached 92.5%. The key contribution of this study is effective accuracy as well as we have trained the model on our own prepared dataset that we have prepared from real world environment. Data Availability: https://drive.google.com/file/d/1bD7B257l-6wqUCQHBWhle95xyrotUbwO/view?usp=sharing\n_____\n\n_____\n**[Fine-Grained Plant Identification using wide and deep learning model 1](https://doi.org/10.1109/PLATCON.2019.8669407)**\n\n\n> **TL;DR:** We propose a model to address the fine-grained plant image classification task by using the wide and deep learning framework which combines a linear model and a deep learning model. Proposed method sums the result of the wide and deep learning model using a logistic function so that discrete features can be considered simultaneously with continuous image content.\n\nIn recent years, with the evolution of deep learning technology, the performance of plant image recognition has improved remarkably. In this paper, we propose a model to address the fine-grained plant image classification task by using the wide and deep learning framework which combines a linear model and a deep learning model. Proposed method sums the result of the wide and deep learning model using a logistic function so that discrete features can be considered simultaneously with continuous image content. Our works used metadata such as the date of flowering and locational information for the wide model. Our experiment shows that the proposed method gives better performance than a baseline method.\n_____\n\n_____\n**[Plant Identification: Experts vs. Machines in the Era of Deep Learning - Deep Learning Techniques Challenge Flora Experts](https://doi.org/10.1007/978-3-319-76445-0_8)**\n\n\n> **TL;DR:** Nine deep learning models were evaluated with regard to nine French botanists. The performance of the deep learning models was found to be close to the human expertise.\n\nAutomated identification of plants and animals have improved considerably in the last few years, in particular thanks to the recent advances in deep learning. The next big question is how far such automated systems are from the human expertise. Indeed, even the best experts are sometimes confused and/or disagree between each others when validating visual or audio observations of living organism. A picture or a sound actually contains only a partial information that is usually not sufficient to determine the right species with certainty. Quantifying this uncertainty and comparing it to the performance of automated systems is of high interest for both computer scientists and expert naturalists. This chapter reports an experimental study following this idea in the plant domain. In total, nine deep-learning systems implemented by three different research teams were evaluated with regard to nine expert botanists of the French flora. Therefore, we created a small set of plant observations that were identified in the field and revised by experts in order to have a near-perfect golden standard. The main outcome of this work is that the performance of state-of-the-art deep learning models is now close to the most advanced human expertise. This shows that automated plant identification systems are now mature enough for several routine tasks, and can offer very promising tools for autonomous ecological surveillance systems.\n_____\n\n_____\n**[Observation on Therapeutic Plant Identification based on Deep Learning Technique](https://doi.org/10.35940/ijitee.j9808.0981119)**\n\n\n> **TL;DR:** There are many different methods for leaf classification, but the most successful ones rely on deep learning algorithms. These algorithms can learn to identify patterns in data that are too complex for humans to discern. This makes them very effective at identifying different types of leaves.\n\nPlants have been used for medicinal purposes long before recorded history. It plays a major role in medicines, food, perfumes and cosmetics industries. By knowing the herbal plants and its usage it can be used for above applications. In this digital era, people don’t have adequate knowledge to identify various herbal plants which are used by our ancestors for long time. Presently, the identification of herbal plants is purely based on the human perception or knowledge. There may be probability of human error occurring. In order to have an efficient herb species classification, there must be a complete model which should be automatic and convenient recognition system. This paper is reviewing the different leaf classification methodologies based on deep learning algorithms. The main aim of this research paper is to conclude the advanced technique for the leaf identification.\n_____\n\n_____\n**[Deep learning for plant identification: how the web can compete with human experts](https://doi.org/10.3897/BISS.2.25637)**\n\nAutomated identification of plants and animals has improved considerably in the last few years, in particular thanks to the recent advances in deep learning. In order to evaluate the performance of automated plant identification technologies in a sustainable and repeatable way, a dedicated system-oriented benchmark was setup in 2011 in the context of ImageCLEF (Goëau et al. 2011). Each year, since that time, several research groups participated in this large collaborative evaluation by benchmarking their image-based plant identification systems. In 2014, the LifeCLEF research platform (Joly et al. 2014) was created in the continuity of this effort so as to enlarge the evaluated challenges by considering birds and fishes in addition to plants, and audio and video contents in addition to images. The 2017-th edition of the LifeCLEF plant identification challenge (Joly et al. 2017) is an important milestone towards automated plant identification systems working at the scale of continental floras with 10.000 plant species living mainly in Europe and North America ‡,§ | §,‡ ¶ # ¤ © Goëau H et al. This is an open access article distributed under the terms of the Creative Commons Attribution License (CC BY 4.0), which permits unrestricted use, distribution, and reproduction in any medium, provided the original author and source are credited. illustrated by a total of 1.1M images. Nowadays, such ambitious systems are enabled thanks to the conjunction of the dazzling recent progress in image classification with deep learning and several outstanding international initiatives, aggregating the visual knowledge on plant species coming from the main national botanical institutes. The PlantCLEF plant challenge that we propose to present at this workshop aimed at evaluating to what extent a large noisy training dataset collected through the web (then containing a lot of labelling errors) can compete with a smaller but trusted training dataset checked by experts. To fairly compare both training strategies, the test dataset was created from a third data source, the Pl@ntNet (Joly et al. 2015) mobile application that collects millions of plant image queries all over the world. Due to the good results obtained at the 2017-th edition of the LifeCLEF plant identification challenge, the next big question is how far such automated systems are from the human expertise. Indeed, even the best experts are sometimes confused and/or disagree with each other when validating images of living organism. A multimedia data actually contains only partial information that is usually not sufficient to determine the right species with certainty. Quantifying this uncertainty and comparing it to the performance of automated systems is of high interest for both computer scientists and expert naturalists. This work reports an experimental study following this idea in the plant domain. In total, 9 deeplearning systems implemented by 3 different research teams were evaluated with regard to 9 expert botanists of the French flora. The main outcome of this work is that the performance of state-of-the-art deep learning models is now close to the most advanced human expertise. This shows that automated plant identification systems are now mature enough for several routine tasks, and can offer very promising tools for autonomous ecological surveillance systems.\n_____\n\n_____\n**[Plant Identification with Deep Learning Ensembles](https://scholar.google.com/scholar?q=Plant%20Identification%20with%20Deep%20Learning%20Ensembles)**\n\nThis work describes the plant identification system that we submitted to the ExpertLifeCLEF plant identification campaign in 2018. We fine-tuned two pre-trained deep learning architectures (SeNet and DensNetwork) using images shared by the CLEF organizers in 2017. Our main runs are 4 ensembles obtained with different weighted combinations of the 4 deep learning architectures. The fifth ensemble is based on deep learning features but uses Error Correcting Output Codes (ECOC) as the ensemble. Our best system has achieved a classification accuracy of 74.4%, while the best system obtained 86.7% accuracy, on the whole of the official test data. This system ranked 4th place among all the teams, but matched the accuracy of one of the human experts.\n_____\n\n_____\n**[A Combination of Deep Learning and Hand-Designed Feature for Plant Identification Based on Leaf and Flower Images](https://doi.org/10.1007/978-3-319-56660-3_20)**\n\n\n> **TL;DR:** This paper proposes a combination of deep learning and hand-designed feature for plant identification based on leaf and flower images. The contributions of this paper are two-fold. First, for each organ image, we have performed a comparative evaluation of deep learning and hand-designed feature for plant identification. Two approaches for deep learning and hand-designed feature that are convolutional neuron network (CNN) and kernel descriptor (KDES) are chosen in our experiments. Second, based on the results of the first contribution, we propose a method for plant identification by late fusing the identification results of leaf and flower. Experimental results on ImageCle\n\nThis paper proposes a combination of deep learning and hand-designed feature for plant identification based on leaf and flower images. The contributions of this paper are two-fold. First, for each organ image, we have performed a comparative evaluation of deep learning and hand-designed feature for plant identification. Two approaches for deep learning and hand-designed feature that are convolutional neuron network (CNN) and kernel descriptor (KDES) are chosen in our experiments. Second, based on the results of the first contribution, we propose a method for plant identification by late fusing the identification results of leaf and flower. Experimental results on ImageClef 2015 dataset show that hand designed feature outperforms deep learning for well-constrained cases (leaf captured on simple background). However, deep learning shows its robustness in natural situations. Moreover, the combination of leaf and flower images improves significantly the identification when comparing leaf-based plant identification.\n_____\n\n_____\n**[Transfer learning with deep convolutional neural network for automated plant identification in hainan island](https://doi.org/10.3233/JIFS-189905)**\n\nThe Hainan Island has a generally high biological diversity with a wide variety of plant species, some of which are listed as endemic to the island. It is time-consuming and difficult, even for the botanist experts to determine the name of species based on observations. Automated plant identification enables experts to process significantly greater numbers of plants with higher efficiencies in shorter periods of time. However, plant recognition is a kind of fine-grained visual recognition problem, which is relatively harder than conventional image recognition. In this paper, we employ a Deep Convolutional Neural Network (DCNN) trained on the ImageNet database, which contains millions of images, and then transfer the learning information for automated plant identification based on flower and fruit images. First, we modify the last three layers of the pre-trained network in order to adapt ResNet-50 model to our classification task, and replace the fully connected layer in the original pre-trained network with another fully connected layers, in which the output size represents the class of plants. Secondly, we use transfer experience and fine-tuned pre-trained DCNN for experiments using flower and fruit images. Finally, we evaluate the proposed network on two available botanical datasets: the Oxford flowers dataset with 102 classes and the HNPlant flowers and fruits dataset with 20 classes, and determine the optimal values of the associated hyperparameters to improve the overall performance. Experiment results demonstrate that the highest classification accuracies exhibited by the proposed model on the Oxford-102 and HNPlant-20 datasets are 92.4% and 95.0%, respectively, thus establishing their effectiveness and superiority.\n_____\n\n_____\n**[Plant diseases and pests detection based on deep learning: a review](https://doi.org/10.1186/s13007-021-00722-9)**\n\nPlant diseases and pests are important factors determining the yield and quality of plants. Plant diseases and pests identification can be carried out by means of digital image processing. In recent years, deep learning has made breakthroughs in the field of digital image processing, far superior to traditional methods. How to use deep learning technology to study plant diseases and pests identification has become a research issue of great concern to researchers. This review provides a definition of plant diseases and pests detection problem, puts forward a comparison with traditional plant diseases and pests detection methods. According to the difference of network structure, this study outlines the research on plant diseases and pests detection based on deep learning in recent years from three aspects of classification network, detection network and segmentation network, and the advantages and disadvantages of each method are summarized. Common datasets are introduced, and the performance of existing studies is compared. On this basis, this study discusses possible challenges in practical applications of plant diseases and pests detection based on deep learning. In addition, possible solutions and research ideas are proposed for the challenges, and several suggestions are given. Finally, this study gives the analysis and prospect of the future trend of plant diseases and pests detection based on deep learning.\n_____\n\n_____\n**[Extended application of deep learning combined with 2DCOS: Study on origin identification in the medicinal plant of Paris polyphylla var. yunnanensis.](https://doi.org/10.1002/pca.3076)**\n\nINTRODUCTION\nMedicinal plants are very important to human health, and ensuring their quality and rapid evaluation are the current research concerns. Deep learning has a strong ability in recognition. This study extended it to the identification of medicinal plants from the perspective of spectrum.\n\n\nOBJECTIVE\nIn order to realise the rapid identification and provide a reference for the selection of high-quality resources of medicinal plants, a combination of deep learning and two-dimensional correlation spectroscopy (2DCOS) was proposed.\n\n\nMETHODS\nFor the first time, Fourier transform mid-infrared (FT-MIR) and near-infrared (NIR) spectroscopy 2DCOS images combined with residual neural network (ResNet) was used for the origin identification of Paris polyphylla var. yunnanensis. In total 1593 samples were collected and 12821 2DCOS images were drawn. The climate of different origins was briefly analysed.\n\n\nRESULTS\nThe xishuangbanna, puer, lincang, honghe and wenshan are the five regions with more ecological advantages. The synchronous 2DCOS models of FT-MIR and NIR could realise origin identification with the accuracy of 100%. The synchronous images were suitable for the identification of medicinal plants with complex systems. The full band, feature band and different contour models had no big difference in distinguishing ability, so they were not the key factors affecting the discrimination results.\n\n\nCONCLUSION\nThe ResNet models established were stable, reliable, and robust, which not only solved the problem of origin identification, expanded the application field of deep learning, but also provided practical reference for the related research of other medicinal plants.\n_____\n\n_____\n**[Plant Disease Identification Using Deep Learning: A Systematic Review](https://doi.org/10.1109/ICIEM51511.2021.9445277)**\n\nAgriculture contributes majorly in the Indian economy being the most important aspect of it. Often the plants suffer from many diseases which may be dependent on climatic conditions and pests which further degrades the quality of the crop. Early and accurate detection of disease in plants is very crucial step in depicting the overall yield of the crop, as this can increase the yield and productivity of the crop by a great margin. In the current climatic conditions, to obtain the superior quality crop is getting difficult day by day as the plants suffer from different diseases. To solve the problem of early and accurate detection image processing has come up with various techniques to find best and suitable ways. This paper presents a review to examine the power of these techniques in detection for plant diseases and add in the agriculture advancement. This survey enfolds larger scope of deep learning in the future research while detecting plant diseases along with improvised performance and accuracy. This paper also states certain challenges which still exist in the detection of diseases in plants and opens some areas of research for the researchers. Various problems in the field of dataset collection have been addressed. Some possible solutions also have been suggested which can help the accuracy to increase for a model.\n_____\n\n_____\n**[Plant Diseases Identification through a Discount Momentum Optimizer in Deep Learning](https://doi.org/10.3390/app11209468)**\n\nDeep learning proves its promising results in various domains. The automatic identification of plant diseases with deep convolutional neural networks attracts a lot of attention at present. This article extends stochastic gradient descent momentum optimizer and presents a discount momentum (DM) deep learning optimizer for plant diseases identification. To examine the recognition and generalization capability of the DM optimizer, we discuss the hyper-parameter tuning and convolutional neural networks models across the plantvillage dataset. We further conduct comparison experiments on popular non-adaptive learning rate methods. The proposed approach achieves an average validation accuracy of no less than 97% for plant diseases prediction on several state-of-the-art deep learning models and holds a low sensitivity to hyper-parameter settings. Experimental results demonstrate that the DM method can bring a higher identification performance, while still maintaining a competitive performance over other non-adaptive learning rate methods in terms of both training speed and generalization.\n_____\n\n_____\n**[A Study of Adversarial Attacks and Detection on Deep Learning-Based Plant Disease Identification](https://doi.org/10.3390/APP11041878)**\n\n\n> **TL;DR:** We found that successful adversarial attacks on deep learning (DL)-based plant disease identification systems could result in a significant delay of treatments and huge economic losses. Adversarial attacks with a small number of perturbations can dramatically degrade the performance of DNN models for plant disease identification. Adversarial attacks can be effectively defended by using adversarial sample detection with an appropriate choice of features.\n\nTransfer learning using pre-trained deep neural networks (DNNs) has been widely used for plant disease identification recently. However, pre-trained DNNs are susceptible to adversarial attacks which generate adversarial samples causing DNN models to make wrong predictions. Successful adversarial attacks on deep learning (DL)-based plant disease identification systems could result in a significant delay of treatments and huge economic losses. This paper is the first attempt to study adversarial attacks and detection on DL-based plant disease identification. Our results show that adversarial attacks with a small number of perturbations can dramatically degrade the performance of DNN models for plant disease identification. We also find that adversarial attacks can be effectively defended by using adversarial sample detection with an appropriate choice of features. Our work will serve as a basis for developing more robust DNN models for plant disease identification and guiding the defense against adversarial attacks.\n_____\n\n_____\n**[Black Measles Disease Identification in Grape Plant (Vitis vinifera) Using Deep Learning](https://doi.org/10.1109/ICCCIS51004.2021.9397205)**\n\nThe most common diseases found in plants are the fungi infections/diseases. One of the common fungal diseases is Esca (Black Measles) which is found in the Grape Plants and can be easily identified as brown streaking lesions on any part of the leaf. The affected leaves can dry off completely and fall off from the plant prematurely which eventually results in death of the plant. In this work, an improved technique based on Deep Learning algorithm for identifying Esca Black measles in GrapeVines is proposed. The proposed method yields better performance and accuracy in detecting the disease, than past Machine Learning based approaches. Grape Plant dataset from PlantVillage Database is used for the work. The dataset contains total 1807 images (healthy and diseased). ResNet 50 architecture of Deep Neaural Network in combination with Transfer Learning and Fine Tuning was used to compute the results. The proposed system provides an accuracy of more than 97% and performed better than the existing approaches which are based on feature extraction methods.\n_____\n\n_____\n**[Identification of Plant Stomata Based on YOLO v5 Deep Learning Model](https://doi.org/10.1145/3507548.3507560)**\n\nStomata is an important structure in all terrestrial plants and is very vital in controlling plant photosynthesis and transpiration flow. Precise detection of plant stomata is the basis for studying stomata characteristics. Traditional detection methods are mostly manual operations, which is a tedious and inefficient process. Manually extracting features requires high image quality. Choosing appropriate features depends on certain prior knowledge, especially for the object with large morphological changes such as plant stomata. With the widespread use of deep learning technology, efficient solutions to this task have become possible. This article combines the characteristics of the corn leaf stomatal data sets to improve the latest object detection model YOLO v5)You Only Look Once(. By introducing the attention mechanism, that is, adding the SE module to the backbone network, the precision and recall of stoma detection are improved. At the same time, The loss function has been improved from to for avoiding some problems that may occur when selecting the best prediction box. Experimental results show that the precision and recall rates of the improved model on the corn leaf stomata data sets have reached 94.8% and 98.7% respectively, lay the foundation for the measurement of stomatal parameters. In addition, this paper also can help agriculturists and botanists to build their own data sets for stomatal research by explaining the methods of acquiring, pre-processing, and annotating data sets.\n_____\n\n_____\n**[Detection of plant leaf diseases using recent progress in Deep Learning-Based identification techniques](https://doi.org/10.5935/JETIA.V7I30.768)**\n\nMostly economy profoundly depends on farming efficiency. The farming crops are commonly affected by the disease. Since the economy depends on agriculture, this is one of the core reasons that infection identification in plants assumes a significant job in the horticulture field. On the off chance that legitimate consideration isn't taken here, at that point, it causes natural consequences for plants and because of which particular item quality, amount, or efficiency are influence. Crop misfortune because of ailments considerably influences the economy and undermines food accessibility. Quick and precise plant ailment location is essential to expanding farming efficiency in a supportable manner. In any case, plant location by human specialists is costly, tedious, and sometimes unrealistic. To counter these difficulties, Plant pathologists want an exact and dependable plant sickness conclusion framework. The on-going utilization of deep learning procedure with image processing methods for plant sickness acknowledgment has become a hot examination subject to give programmed analysis. This research provides a productive plant illness distinguishing proof technique dependent on pre-prepared deep learning models, such as AlexNet and GoogleNet designs. We trust that this work will be a significant asset for analysts in the area of ailment acknowledgment utilizing image handling strategies with deep learning architectures.\n_____\n\n_____\n**[Identification of Rice Plant Diseases Using Image Processing, Machine Learning & Deep Learning: A Review](https://scholar.google.com/scholar?q=Identification%20of%20Rice%20Plant%20Diseases%20Using%20Image%20Processing%2C%20Machine%20Learning%20%26%20Deep%20Learning%3A%20A%20Review)**\n\n\n> **TL;DR:** This article discusses different methods for detecting rice plant diseases, with a focus on deep learning methods. It is found that deep learning methods are more promising than other methods for this task.\n\nAgriculture is the primary source of livelihood for about more than 50% of the Indian population and rice is one of the major food grains of India. It is observed that rice plant diseases are the major contributors to reduce the production & quality of food. Identification of such diseases may improve the production quality. This paper gives an idea about different methods such as image processing, machine learning & deep learning which are used to detect deadly diseases in rice plants. Much research has been done to automate the rice plant disease detection process using images of the leaf. This manuscript has compared different rice plant disease detection methods and it is found that deep learning methods are more promising than other two methods.\n_____\n\n_____\n**[Plant Disease Identification on Real-World Data using Deep Learning: A Comparative Study](https://doi.org/10.1109/ICCCT53315.2021.9711781)**\n\nCorrect and timely identification of disease in plants, especially for a country like India, where agriculture continues to serve as a cornerstone of its economy, is an indispensable tool demanding our solicitousness. Recently, researchers have started to employ autonomous real-time systems involving deep learning techniques for this purpose. However, the wide va-riety of heterogeneous diseases affecting crop yield continues to prove itself a mammoth task for farmers and stymies the researchers. The majority of the current state-of-the-art models utilize datasets like Plant Village, consisting of leaf images taken in a controlled lab environment, which do not serve as accurate representative data of the real-world scenario. Moreover, the effectiveness of state-of-the-art models like EfficientNetLite, which provided notable improvements in accuracy for similar deep learning applications, remains untested on plant disease datasets. Hence, an exhaustive study on the performance of various state-of-the-art models with varied training conditions on real-time datasets like PlantDoc is imperative to further this area of research. In this paper, we have explored state-of-the-art CNNs like InceptionResNet, EfficientNetLite_0, and VGG-19, under various parameter settings, image augmentations techniques, and loss functions on the real-time PlantDoc dataset. We have evaluated and presented a comparative analysis of the exhaustive combinations and performances gauged by accuracy, top-5 accuracy, F1 scores, and other inferences drawn from training and testing all these networks on 27 different classes of crop diseases. We infer that EfficientNetLite proved to be a most effective architecture, especially given its relatively smaller size. EfficientNetLite coupled with the focal loss function and Albumentaions augmentation library yielded the best results with an accuracy of 70.71%, mean F1-score as 0.7, and 95.57% top-5 accuracy, on a test set congruent with real-world relevance.\n_____\n\n_____\n**[Ensemble Deep Learning Models for Fine-grained Plant Species Identification](https://doi.org/10.1109/CSDE53843.2021.9718387)**\n\n\n> **TL;DR:** This study investigates the development and application of ensemble deep learning models for fine-grained plant species identification. Six popular deep learning models (InceptionResNetV2, ResNet50, Xception, InceptionV3, MobileNetV2, and GoogleNet) were trained on both datasets and heterogeneous ensembles with various ensemble techniques (mean, weighted mean, voting, and stacked generalization) were performed. The validation and testing accuracy results for individual models were compared with the output generated by the ensemble methods.\n\nAutomated plant species identification for the datasets (images) collected from the natural environment is a challenging task. This study investigates the development and application of ensemble deep learning models for fine-grained plant species identification. Two different types of plant species datasets have been used in this study. The first dataset (UBD_45) consists of 45 medicinal plant species from the natural environment with the imbalanced distribution of classes and the second dataset (VP_200) has 200 medicinal plant species with balanced classes from the natural environment. Six popular deep learning models (InceptionResNetV2, ResNet50, Xception, InceptionV3, MobileNetV2, and GoogleNet) were trained on both datasets and heterogeneous ensembles with various ensemble techniques (mean, weighted mean, voting, and stacked generalization) were performed. The validation and testing accuracy results for individual models were compared with the output generated by the ensemble methods. The highest testing accuracies for base models were found 96.7% and 91.2% for UBD_45 and VP_200 datasets, respectively. Mean, weighted mean, and stacking ensembles showed better performance for both datasets. The stacking ensemble improved the classification accuracy by around 1.8% for the UBD_45 dataset while for VP_200 a significant improvement of around 4.23% was noticed using a weighted mean ensemble.\n_____\n\n_____\n**[A Deep Learning Approach for Plant Material Disease Identification](https://doi.org/10.1088/1757-899X/1116/1/012133)**\n\nPlant Material Disease Identification is essential for the food safety. To increase the crop production for the growing population of the world, the proper treatment is required on proper time to save the plant. Therefore, disease diagnosing on time is very important. This paper uses a deep learning convolutional neural network model to identify the plant disease. The pre-existing deep learning model Alexnet has been employed for plant disease identification in which an external feature of segmented plant material (leaves) is passed to the deepest fully connected layer. This combination of extracted feature by Alexnet and external feature of segmented plant material helps in plant disease identification. Experimental analysis has been done on a standard dataset Plant Village which has total 54,306 leaf images of 15 distinct plants having 38 diseases. The presented CNN approach worked well and outperformed to the existing approach.\n_____\n\n_____\n**[Plant Disease Identification Based on Deep Learning Algorithm in Smart Farming](https://doi.org/10.1155/2020/2479172)**\n\nThe identification of plant disease is the premise of the prevention of plant disease efficiently and precisely in the complex environment. With the rapid development of the smart farming, the identification of plant disease becomes digitalized and data-driven, enabling advanced decision support, smart analyses, and planning. This paper proposes a mathematical model of plant disease detection and recognition based on deep learning, which improves accuracy, generality, and training efficiency. Firstly, the region proposal network (RPN) is utilized to recognize and localize the leaves in complex surroundings. Then, images segmented based on the results of RPN algorithm contain the feature of symptoms through Chan–Vese (CV) algorithm. Finally, the segmented leaves are input into the transfer learning model and trained by the dataset of diseased leaves under simple background. Furthermore, the model is examined with black rot, bacterial plaque, and rust diseases. The results show that the accuracy of the method is 83.57%, which is better than the traditional method, thus reducing the influence of disease on agricultural production and being favorable to sustainable development of agriculture. Therefore, the deep learning algorithm proposed in the paper is of great significance in intelligent agriculture, ecological protection, and agricultural production.\n_____\n", "votes": 5 }, { "id": 1778954, "postDate": "2022-05-06T00:22:41.343Z", "content": "

Hi @satoshidatamoto! Thanks for sharing this important piece of information!

", "rawMarkdown": "Hi @satoshidatamoto! Thanks for sharing this important piece of information!" } ], "comments": [ { "id": 1778954, "author_name": "John Park", "author_url": "", "post_date": "2022-05-06T00:22:41.343000", "content": "

Hi @satoshidatamoto! Thanks for sharing this important piece of information!

", "votes": 0, "replies": [] } ], "raw_markdown_by_id": { "1770551": "\nThese are the papers that will change your life.\n\nThese are the papers that will make you question everything you know.\n\nThese are the papers that will make you see the world in a whole new way.\n\nThese are the papers that will give you hints on improving your LB score ;)\n\nHave fun!\n\n_____\n**[Medicinal Plant Identification using Gabor Filters and Deep Learning Techniques: A Paper Review](https://doi.org/10.3844/jcssp.2021.1210.1221)**\n\n\n> **TL;DR:** Computer-aided identification of plants is a branch of machine learning that has become more recognized recently and proves itself as a vital tool in numerous sectors including pharmacological science, forestry and agriculture. This study reviewed plant species classification relying on leaf textural features using Gabor filters and revealed that Gabor filters perform better when combined with other feature extraction methods. Therefore, this study proposes using Log-Gabor filter in the field of plant identification to improve accuracy since they overcome the drawbacks of Gabor filters which are; the maximum bandwidth of a Gabor filter is limited to approximately one octave and Gabor filters are not optimal if one\n\nCorresponding Author: Stephen Opoku Oppong ICT Education Department, University of Education, Winneba, Ghana Email: sooppong@uew.edu.gh Abstract: Computer-aided identification of plants is a branch of machine learning that has become more recognized recently and proves itself as a vital tool in numerous sectors including pharmacological science, forestry and agriculture. This has essentially generated a zeal in creating automated systems for the identification of diverse species of plants. This study reviewed plant species classification relying on leaf textural features using Gabor filters and revealed that Gabor filters perform better when combined with other feature extraction methods. Therefore, this study proposes using Log-Gabor filter in the field of plant identification to improve accuracy since they overcome the drawbacks of Gabor filters which are; the maximum bandwidth of a Gabor filter is limited to approximately one octave and Gabor filters are not optimal if one is seeking broad spectral information with maximal spatial localization.\n_____\n\n_____\n**[Holistic Based Plant Identification Using Deep Learning](https://doi.org/10.1109/icet54505.2021.9689804)**\n\n\n> **TL;DR:** This paper introduces an automated plant classification framework for recognizing plants dependent on their holistic shape. The system uses image preprocessing, deep learning techniques, and ResNet18 to identify plants through the camera; it will give their names in Urdu, English, and the plant's scientific name as an output. Moreover, the system can describe the plant along with its audio. The methodology was evaluated using two datasets. The first dataset, GRANDYMU, consists of 70 different plant species containing 3500 images, and the second is the famous SWEDISH leaf dataset. We trained different models based on Resnet18\n\nPlants are the most important forms of life on earth. They are the fundamental asset for human prosperity, furnishing us with oxygen and food. Humans have widely utilized plants in medication, food formation, and the cosmetic business worldwide in their daily life activities. Identification of plants is extremely a challenging task due to the recursive nature of the shapes of plants. Moreover, identification has been done on a few categories and is limited to veins or leaf patterns. This paper introduced an automated plant classification framework for recognizing plants dependent on their holistic shape. The system uses image preprocessing, deep learning techniques, and ResNet18 to identify plants through the camera; it will give their names in Urdu, English, and the plant's scientific name as an output. Moreover, the system can describe the plant along with its audio. The methodology was evaluated using two datasets. The first dataset, GRANDYMU, consists of 70 different plant species containing 3500 images, and the second is the famous SWEDISH leaf dataset. We trained different models based on Resnet18, and the maximum accuracy that we achieved was 99%. Also, we got an accuracy of 99.9% for the latter dataset.\n_____\n\n_____\n**[Experimental evaluation of Data Augmentation heuristics for plant identification systems based on Deep Learning](https://doi.org/10.5753/sbiagro.2021.18384)**\n\nData augmentation (DA) allows increasing datasets for training machine learning models that demands large amounts of data. In real-world applications in which data may not be abundant enough and data acquisition is not easy, DA enables increasing diversity and introducing model generalization. In this work we evaluate several DA techniques and combining approaches to extend image datasets used to train plant species recognition models. We experimentally validated Deep Convolutional Neural Networks (DCNN) with several datasets obtained from common augmentation techniques and combinations. The results allowed the identification of the Translate + Crop augmentation policy as the most effective within the scope of evaluation.\n_____\n\n_____\n**[Deep Learning for Plant Identification in Natural Environment](https://doi.org/10.1155/2017/7361042)**\n\n\n> **TL;DR:** We collected a dataset of 10,000 images of 100 ornamental plant species in Beijing Forestry University campus using mobile phones. A 26-layer deep learning model consisting of 8 residual building blocks was designed for large-scale plant classification in natural environment. The proposed model achieved a recognition rate of 91.78% on the BJFU100 dataset, demonstrating that deep learning is a promising technology for smart forestry.\n\nPlant image identification has become an interdisciplinary focus in both botanical taxonomy and computer vision. The first plant image dataset collected by mobile phone in natural scene is presented, which contains 10,000 images of 100 ornamental plant species in Beijing Forestry University campus. A 26-layer deep learning model consisting of 8 residual building blocks is designed for large-scale plant classification in natural environment. The proposed model achieves a recognition rate of 91.78% on the BJFU100 dataset, demonstrating that deep learning is a promising technology for smart forestry.\n_____\n\n_____\n**[Plant Identification Based on Noisy Web Data: the Amazing Performance of Deep Learning (LifeCLEF 2017)](https://scholar.google.com/scholar?q=Plant%20Identification%20Based%20on%20Noisy%20Web%20Data%3A%20the%20Amazing%20Performance%20of%20Deep%20Learning%20(LifeCLEF%202017))**\n\nThe 2017 fith edition of the LifeCLEF plant identification challenge is an important milestone towards automated plant identification systems working at the scale of continental floras with 10.000 plant species living mainly in Europe and North America illustrated by a total of 1.1M images. Nowadays, such ambitious systems are enabled thanks to the conjunction of the dazzling recent progress in image classification with deep learning and several outstanding international initiatives, such as the Encyclopedia of Life (EOL), aggregating the visual knowledge on plant species coming from the main national botany institutes. However, despite all these efforts the majority of the plant species still remain without pictures or are poorly illustrated. Outside the institutional channels, a much larger number of plant pictures are available and spread on the web through botanist blogs, plant lovers web-pages, image hosting websites and on-line plant retailers. The LifeCLEF 2017 plant challenge presented in this paper aimed at evaluating to what extent a large noisy training dataset collected through the web and containing a lot of labelling errors can compete with a smaller but trusted training dataset checked by experts. To fairly compare both training strategies, the test dataset was created from a third data source, i.e. the Pl@ntNet mobile application that collects millions of plant image queries all over the world. This paper presents more precisely the resources and assessments of the challenge, summarizes the approaches and systems employed by the participating research groups, and provides an analysis of the main outcomes. (Resume d'auteur)\n_____\n\n_____\n**[Deep Learning with Taxonomic Loss for Plant Identification](https://doi.org/10.1155/2019/2015017)**\n\n\n> **TL;DR:** Train a neural network on a plant identification dataset using a taxonomic loss function to improve performance.\n\nPlant identification is a fine-grained classification task which aims to identify the family, genus, and species according to plant appearance features. Inspired by the hierarchical structure of taxonomic tree, the taxonomic loss was proposed, which could encode the hierarchical relationships among multilevel labels into the deep learning objective function by simple group and sum operation. By training various neural networks on PlantCLEF 2015 and PlantCLEF 2017 datasets, the experimental results demonstrated that the proposed loss function was easy to implement and outperformed the most commonly adopted cross-entropy loss. Eight neural networks were trained, respectively, by two different loss functions on PlantCLEF 2015 dataset, and the models trained by taxonomic loss led to significant performance improvements. On PlantCLEF 2017 dataset with 10,000 species, the SENet-154 model trained by taxonomic loss achieved the accuracies of 84.07%, 79.97%, and 73.61% at family, genus and species levels, which improved those of model trained by cross-entropy loss by 2.23%, 1.34%, and 1.08%, respectively. The taxonomic loss could further facilitate the fine-grained classification task with hierarchical labels.\n_____\n\n_____\n**[A Light Weight Deep Learning Model for Real World Plant Identification](https://doi.org/10.1109/dchpc55044.2022.9731841)**\n\nAutomatic identification and classification of different plant leaf species have become a common trend among researchers and scientists. To obtain a result with better precision, they use various methods and techniques of deep learning to build a model. Convolutional neural networks are becoming the most common method used by scientists to classify plant leaves. However, the classification of plant leaves can be challenging with more rare species and complicated backgrounds, for which researchers build several models to achieve high-level accuracy. In the present study for the classification of leaves, we have created a model for plant leaf classification based on a dataset we collected. We've used the Resnet-50 model, a well-known CNN architecture, which provided an efficient method to organize and analyze a deep classification to reduce the complexity so that there will be fewer parameters for training and low time consumption as well. Using Resnet-50, we intended to develop a significant result in our classification model. The convolutional neural network is famous for its influential abilities in feature extraction and classification. And Resnet-50 being a residual network enabled us to train deep networks in our model. The average training accuracy reached 98.3%, while the average testing accuracy reached 92.5%. The key contribution of this study is effective accuracy as well as we have trained the model on our own prepared dataset that we have prepared from real world environment. Data Availability: https://drive.google.com/file/d/1bD7B257l-6wqUCQHBWhle95xyrotUbwO/view?usp=sharing\n_____\n\n_____\n**[Fine-Grained Plant Identification using wide and deep learning model 1](https://doi.org/10.1109/PLATCON.2019.8669407)**\n\n\n> **TL;DR:** We propose a model to address the fine-grained plant image classification task by using the wide and deep learning framework which combines a linear model and a deep learning model. Proposed method sums the result of the wide and deep learning model using a logistic function so that discrete features can be considered simultaneously with continuous image content.\n\nIn recent years, with the evolution of deep learning technology, the performance of plant image recognition has improved remarkably. In this paper, we propose a model to address the fine-grained plant image classification task by using the wide and deep learning framework which combines a linear model and a deep learning model. Proposed method sums the result of the wide and deep learning model using a logistic function so that discrete features can be considered simultaneously with continuous image content. Our works used metadata such as the date of flowering and locational information for the wide model. Our experiment shows that the proposed method gives better performance than a baseline method.\n_____\n\n_____\n**[Plant Identification: Experts vs. Machines in the Era of Deep Learning - Deep Learning Techniques Challenge Flora Experts](https://doi.org/10.1007/978-3-319-76445-0_8)**\n\n\n> **TL;DR:** Nine deep learning models were evaluated with regard to nine French botanists. The performance of the deep learning models was found to be close to the human expertise.\n\nAutomated identification of plants and animals have improved considerably in the last few years, in particular thanks to the recent advances in deep learning. The next big question is how far such automated systems are from the human expertise. Indeed, even the best experts are sometimes confused and/or disagree between each others when validating visual or audio observations of living organism. A picture or a sound actually contains only a partial information that is usually not sufficient to determine the right species with certainty. Quantifying this uncertainty and comparing it to the performance of automated systems is of high interest for both computer scientists and expert naturalists. This chapter reports an experimental study following this idea in the plant domain. In total, nine deep-learning systems implemented by three different research teams were evaluated with regard to nine expert botanists of the French flora. Therefore, we created a small set of plant observations that were identified in the field and revised by experts in order to have a near-perfect golden standard. The main outcome of this work is that the performance of state-of-the-art deep learning models is now close to the most advanced human expertise. This shows that automated plant identification systems are now mature enough for several routine tasks, and can offer very promising tools for autonomous ecological surveillance systems.\n_____\n\n_____\n**[Observation on Therapeutic Plant Identification based on Deep Learning Technique](https://doi.org/10.35940/ijitee.j9808.0981119)**\n\n\n> **TL;DR:** There are many different methods for leaf classification, but the most successful ones rely on deep learning algorithms. These algorithms can learn to identify patterns in data that are too complex for humans to discern. This makes them very effective at identifying different types of leaves.\n\nPlants have been used for medicinal purposes long before recorded history. It plays a major role in medicines, food, perfumes and cosmetics industries. By knowing the herbal plants and its usage it can be used for above applications. In this digital era, people don’t have adequate knowledge to identify various herbal plants which are used by our ancestors for long time. Presently, the identification of herbal plants is purely based on the human perception or knowledge. There may be probability of human error occurring. In order to have an efficient herb species classification, there must be a complete model which should be automatic and convenient recognition system. This paper is reviewing the different leaf classification methodologies based on deep learning algorithms. The main aim of this research paper is to conclude the advanced technique for the leaf identification.\n_____\n\n_____\n**[Deep learning for plant identification: how the web can compete with human experts](https://doi.org/10.3897/BISS.2.25637)**\n\nAutomated identification of plants and animals has improved considerably in the last few years, in particular thanks to the recent advances in deep learning. In order to evaluate the performance of automated plant identification technologies in a sustainable and repeatable way, a dedicated system-oriented benchmark was setup in 2011 in the context of ImageCLEF (Goëau et al. 2011). Each year, since that time, several research groups participated in this large collaborative evaluation by benchmarking their image-based plant identification systems. In 2014, the LifeCLEF research platform (Joly et al. 2014) was created in the continuity of this effort so as to enlarge the evaluated challenges by considering birds and fishes in addition to plants, and audio and video contents in addition to images. The 2017-th edition of the LifeCLEF plant identification challenge (Joly et al. 2017) is an important milestone towards automated plant identification systems working at the scale of continental floras with 10.000 plant species living mainly in Europe and North America ‡,§ | §,‡ ¶ # ¤ © Goëau H et al. This is an open access article distributed under the terms of the Creative Commons Attribution License (CC BY 4.0), which permits unrestricted use, distribution, and reproduction in any medium, provided the original author and source are credited. illustrated by a total of 1.1M images. Nowadays, such ambitious systems are enabled thanks to the conjunction of the dazzling recent progress in image classification with deep learning and several outstanding international initiatives, aggregating the visual knowledge on plant species coming from the main national botanical institutes. The PlantCLEF plant challenge that we propose to present at this workshop aimed at evaluating to what extent a large noisy training dataset collected through the web (then containing a lot of labelling errors) can compete with a smaller but trusted training dataset checked by experts. To fairly compare both training strategies, the test dataset was created from a third data source, the Pl@ntNet (Joly et al. 2015) mobile application that collects millions of plant image queries all over the world. Due to the good results obtained at the 2017-th edition of the LifeCLEF plant identification challenge, the next big question is how far such automated systems are from the human expertise. Indeed, even the best experts are sometimes confused and/or disagree with each other when validating images of living organism. A multimedia data actually contains only partial information that is usually not sufficient to determine the right species with certainty. Quantifying this uncertainty and comparing it to the performance of automated systems is of high interest for both computer scientists and expert naturalists. This work reports an experimental study following this idea in the plant domain. In total, 9 deeplearning systems implemented by 3 different research teams were evaluated with regard to 9 expert botanists of the French flora. The main outcome of this work is that the performance of state-of-the-art deep learning models is now close to the most advanced human expertise. This shows that automated plant identification systems are now mature enough for several routine tasks, and can offer very promising tools for autonomous ecological surveillance systems.\n_____\n\n_____\n**[Plant Identification with Deep Learning Ensembles](https://scholar.google.com/scholar?q=Plant%20Identification%20with%20Deep%20Learning%20Ensembles)**\n\nThis work describes the plant identification system that we submitted to the ExpertLifeCLEF plant identification campaign in 2018. We fine-tuned two pre-trained deep learning architectures (SeNet and DensNetwork) using images shared by the CLEF organizers in 2017. Our main runs are 4 ensembles obtained with different weighted combinations of the 4 deep learning architectures. The fifth ensemble is based on deep learning features but uses Error Correcting Output Codes (ECOC) as the ensemble. Our best system has achieved a classification accuracy of 74.4%, while the best system obtained 86.7% accuracy, on the whole of the official test data. This system ranked 4th place among all the teams, but matched the accuracy of one of the human experts.\n_____\n\n_____\n**[A Combination of Deep Learning and Hand-Designed Feature for Plant Identification Based on Leaf and Flower Images](https://doi.org/10.1007/978-3-319-56660-3_20)**\n\n\n> **TL;DR:** This paper proposes a combination of deep learning and hand-designed feature for plant identification based on leaf and flower images. The contributions of this paper are two-fold. First, for each organ image, we have performed a comparative evaluation of deep learning and hand-designed feature for plant identification. Two approaches for deep learning and hand-designed feature that are convolutional neuron network (CNN) and kernel descriptor (KDES) are chosen in our experiments. Second, based on the results of the first contribution, we propose a method for plant identification by late fusing the identification results of leaf and flower. Experimental results on ImageCle\n\nThis paper proposes a combination of deep learning and hand-designed feature for plant identification based on leaf and flower images. The contributions of this paper are two-fold. First, for each organ image, we have performed a comparative evaluation of deep learning and hand-designed feature for plant identification. Two approaches for deep learning and hand-designed feature that are convolutional neuron network (CNN) and kernel descriptor (KDES) are chosen in our experiments. Second, based on the results of the first contribution, we propose a method for plant identification by late fusing the identification results of leaf and flower. Experimental results on ImageClef 2015 dataset show that hand designed feature outperforms deep learning for well-constrained cases (leaf captured on simple background). However, deep learning shows its robustness in natural situations. Moreover, the combination of leaf and flower images improves significantly the identification when comparing leaf-based plant identification.\n_____\n\n_____\n**[Transfer learning with deep convolutional neural network for automated plant identification in hainan island](https://doi.org/10.3233/JIFS-189905)**\n\nThe Hainan Island has a generally high biological diversity with a wide variety of plant species, some of which are listed as endemic to the island. It is time-consuming and difficult, even for the botanist experts to determine the name of species based on observations. Automated plant identification enables experts to process significantly greater numbers of plants with higher efficiencies in shorter periods of time. However, plant recognition is a kind of fine-grained visual recognition problem, which is relatively harder than conventional image recognition. In this paper, we employ a Deep Convolutional Neural Network (DCNN) trained on the ImageNet database, which contains millions of images, and then transfer the learning information for automated plant identification based on flower and fruit images. First, we modify the last three layers of the pre-trained network in order to adapt ResNet-50 model to our classification task, and replace the fully connected layer in the original pre-trained network with another fully connected layers, in which the output size represents the class of plants. Secondly, we use transfer experience and fine-tuned pre-trained DCNN for experiments using flower and fruit images. Finally, we evaluate the proposed network on two available botanical datasets: the Oxford flowers dataset with 102 classes and the HNPlant flowers and fruits dataset with 20 classes, and determine the optimal values of the associated hyperparameters to improve the overall performance. Experiment results demonstrate that the highest classification accuracies exhibited by the proposed model on the Oxford-102 and HNPlant-20 datasets are 92.4% and 95.0%, respectively, thus establishing their effectiveness and superiority.\n_____\n\n_____\n**[Plant diseases and pests detection based on deep learning: a review](https://doi.org/10.1186/s13007-021-00722-9)**\n\nPlant diseases and pests are important factors determining the yield and quality of plants. Plant diseases and pests identification can be carried out by means of digital image processing. In recent years, deep learning has made breakthroughs in the field of digital image processing, far superior to traditional methods. How to use deep learning technology to study plant diseases and pests identification has become a research issue of great concern to researchers. This review provides a definition of plant diseases and pests detection problem, puts forward a comparison with traditional plant diseases and pests detection methods. According to the difference of network structure, this study outlines the research on plant diseases and pests detection based on deep learning in recent years from three aspects of classification network, detection network and segmentation network, and the advantages and disadvantages of each method are summarized. Common datasets are introduced, and the performance of existing studies is compared. On this basis, this study discusses possible challenges in practical applications of plant diseases and pests detection based on deep learning. In addition, possible solutions and research ideas are proposed for the challenges, and several suggestions are given. Finally, this study gives the analysis and prospect of the future trend of plant diseases and pests detection based on deep learning.\n_____\n\n_____\n**[Extended application of deep learning combined with 2DCOS: Study on origin identification in the medicinal plant of Paris polyphylla var. yunnanensis.](https://doi.org/10.1002/pca.3076)**\n\nINTRODUCTION\nMedicinal plants are very important to human health, and ensuring their quality and rapid evaluation are the current research concerns. Deep learning has a strong ability in recognition. This study extended it to the identification of medicinal plants from the perspective of spectrum.\n\n\nOBJECTIVE\nIn order to realise the rapid identification and provide a reference for the selection of high-quality resources of medicinal plants, a combination of deep learning and two-dimensional correlation spectroscopy (2DCOS) was proposed.\n\n\nMETHODS\nFor the first time, Fourier transform mid-infrared (FT-MIR) and near-infrared (NIR) spectroscopy 2DCOS images combined with residual neural network (ResNet) was used for the origin identification of Paris polyphylla var. yunnanensis. In total 1593 samples were collected and 12821 2DCOS images were drawn. The climate of different origins was briefly analysed.\n\n\nRESULTS\nThe xishuangbanna, puer, lincang, honghe and wenshan are the five regions with more ecological advantages. The synchronous 2DCOS models of FT-MIR and NIR could realise origin identification with the accuracy of 100%. The synchronous images were suitable for the identification of medicinal plants with complex systems. The full band, feature band and different contour models had no big difference in distinguishing ability, so they were not the key factors affecting the discrimination results.\n\n\nCONCLUSION\nThe ResNet models established were stable, reliable, and robust, which not only solved the problem of origin identification, expanded the application field of deep learning, but also provided practical reference for the related research of other medicinal plants.\n_____\n\n_____\n**[Plant Disease Identification Using Deep Learning: A Systematic Review](https://doi.org/10.1109/ICIEM51511.2021.9445277)**\n\nAgriculture contributes majorly in the Indian economy being the most important aspect of it. Often the plants suffer from many diseases which may be dependent on climatic conditions and pests which further degrades the quality of the crop. Early and accurate detection of disease in plants is very crucial step in depicting the overall yield of the crop, as this can increase the yield and productivity of the crop by a great margin. In the current climatic conditions, to obtain the superior quality crop is getting difficult day by day as the plants suffer from different diseases. To solve the problem of early and accurate detection image processing has come up with various techniques to find best and suitable ways. This paper presents a review to examine the power of these techniques in detection for plant diseases and add in the agriculture advancement. This survey enfolds larger scope of deep learning in the future research while detecting plant diseases along with improvised performance and accuracy. This paper also states certain challenges which still exist in the detection of diseases in plants and opens some areas of research for the researchers. Various problems in the field of dataset collection have been addressed. Some possible solutions also have been suggested which can help the accuracy to increase for a model.\n_____\n\n_____\n**[Plant Diseases Identification through a Discount Momentum Optimizer in Deep Learning](https://doi.org/10.3390/app11209468)**\n\nDeep learning proves its promising results in various domains. The automatic identification of plant diseases with deep convolutional neural networks attracts a lot of attention at present. This article extends stochastic gradient descent momentum optimizer and presents a discount momentum (DM) deep learning optimizer for plant diseases identification. To examine the recognition and generalization capability of the DM optimizer, we discuss the hyper-parameter tuning and convolutional neural networks models across the plantvillage dataset. We further conduct comparison experiments on popular non-adaptive learning rate methods. The proposed approach achieves an average validation accuracy of no less than 97% for plant diseases prediction on several state-of-the-art deep learning models and holds a low sensitivity to hyper-parameter settings. Experimental results demonstrate that the DM method can bring a higher identification performance, while still maintaining a competitive performance over other non-adaptive learning rate methods in terms of both training speed and generalization.\n_____\n\n_____\n**[A Study of Adversarial Attacks and Detection on Deep Learning-Based Plant Disease Identification](https://doi.org/10.3390/APP11041878)**\n\n\n> **TL;DR:** We found that successful adversarial attacks on deep learning (DL)-based plant disease identification systems could result in a significant delay of treatments and huge economic losses. Adversarial attacks with a small number of perturbations can dramatically degrade the performance of DNN models for plant disease identification. Adversarial attacks can be effectively defended by using adversarial sample detection with an appropriate choice of features.\n\nTransfer learning using pre-trained deep neural networks (DNNs) has been widely used for plant disease identification recently. However, pre-trained DNNs are susceptible to adversarial attacks which generate adversarial samples causing DNN models to make wrong predictions. Successful adversarial attacks on deep learning (DL)-based plant disease identification systems could result in a significant delay of treatments and huge economic losses. This paper is the first attempt to study adversarial attacks and detection on DL-based plant disease identification. Our results show that adversarial attacks with a small number of perturbations can dramatically degrade the performance of DNN models for plant disease identification. We also find that adversarial attacks can be effectively defended by using adversarial sample detection with an appropriate choice of features. Our work will serve as a basis for developing more robust DNN models for plant disease identification and guiding the defense against adversarial attacks.\n_____\n\n_____\n**[Black Measles Disease Identification in Grape Plant (Vitis vinifera) Using Deep Learning](https://doi.org/10.1109/ICCCIS51004.2021.9397205)**\n\nThe most common diseases found in plants are the fungi infections/diseases. One of the common fungal diseases is Esca (Black Measles) which is found in the Grape Plants and can be easily identified as brown streaking lesions on any part of the leaf. The affected leaves can dry off completely and fall off from the plant prematurely which eventually results in death of the plant. In this work, an improved technique based on Deep Learning algorithm for identifying Esca Black measles in GrapeVines is proposed. The proposed method yields better performance and accuracy in detecting the disease, than past Machine Learning based approaches. Grape Plant dataset from PlantVillage Database is used for the work. The dataset contains total 1807 images (healthy and diseased). ResNet 50 architecture of Deep Neaural Network in combination with Transfer Learning and Fine Tuning was used to compute the results. The proposed system provides an accuracy of more than 97% and performed better than the existing approaches which are based on feature extraction methods.\n_____\n\n_____\n**[Identification of Plant Stomata Based on YOLO v5 Deep Learning Model](https://doi.org/10.1145/3507548.3507560)**\n\nStomata is an important structure in all terrestrial plants and is very vital in controlling plant photosynthesis and transpiration flow. Precise detection of plant stomata is the basis for studying stomata characteristics. Traditional detection methods are mostly manual operations, which is a tedious and inefficient process. Manually extracting features requires high image quality. Choosing appropriate features depends on certain prior knowledge, especially for the object with large morphological changes such as plant stomata. With the widespread use of deep learning technology, efficient solutions to this task have become possible. This article combines the characteristics of the corn leaf stomatal data sets to improve the latest object detection model YOLO v5)You Only Look Once(. By introducing the attention mechanism, that is, adding the SE module to the backbone network, the precision and recall of stoma detection are improved. At the same time, The loss function has been improved from to for avoiding some problems that may occur when selecting the best prediction box. Experimental results show that the precision and recall rates of the improved model on the corn leaf stomata data sets have reached 94.8% and 98.7% respectively, lay the foundation for the measurement of stomatal parameters. In addition, this paper also can help agriculturists and botanists to build their own data sets for stomatal research by explaining the methods of acquiring, pre-processing, and annotating data sets.\n_____\n\n_____\n**[Detection of plant leaf diseases using recent progress in Deep Learning-Based identification techniques](https://doi.org/10.5935/JETIA.V7I30.768)**\n\nMostly economy profoundly depends on farming efficiency. The farming crops are commonly affected by the disease. Since the economy depends on agriculture, this is one of the core reasons that infection identification in plants assumes a significant job in the horticulture field. On the off chance that legitimate consideration isn't taken here, at that point, it causes natural consequences for plants and because of which particular item quality, amount, or efficiency are influence. Crop misfortune because of ailments considerably influences the economy and undermines food accessibility. Quick and precise plant ailment location is essential to expanding farming efficiency in a supportable manner. In any case, plant location by human specialists is costly, tedious, and sometimes unrealistic. To counter these difficulties, Plant pathologists want an exact and dependable plant sickness conclusion framework. The on-going utilization of deep learning procedure with image processing methods for plant sickness acknowledgment has become a hot examination subject to give programmed analysis. This research provides a productive plant illness distinguishing proof technique dependent on pre-prepared deep learning models, such as AlexNet and GoogleNet designs. We trust that this work will be a significant asset for analysts in the area of ailment acknowledgment utilizing image handling strategies with deep learning architectures.\n_____\n\n_____\n**[Identification of Rice Plant Diseases Using Image Processing, Machine Learning & Deep Learning: A Review](https://scholar.google.com/scholar?q=Identification%20of%20Rice%20Plant%20Diseases%20Using%20Image%20Processing%2C%20Machine%20Learning%20%26%20Deep%20Learning%3A%20A%20Review)**\n\n\n> **TL;DR:** This article discusses different methods for detecting rice plant diseases, with a focus on deep learning methods. It is found that deep learning methods are more promising than other methods for this task.\n\nAgriculture is the primary source of livelihood for about more than 50% of the Indian population and rice is one of the major food grains of India. It is observed that rice plant diseases are the major contributors to reduce the production & quality of food. Identification of such diseases may improve the production quality. This paper gives an idea about different methods such as image processing, machine learning & deep learning which are used to detect deadly diseases in rice plants. Much research has been done to automate the rice plant disease detection process using images of the leaf. This manuscript has compared different rice plant disease detection methods and it is found that deep learning methods are more promising than other two methods.\n_____\n\n_____\n**[Plant Disease Identification on Real-World Data using Deep Learning: A Comparative Study](https://doi.org/10.1109/ICCCT53315.2021.9711781)**\n\nCorrect and timely identification of disease in plants, especially for a country like India, where agriculture continues to serve as a cornerstone of its economy, is an indispensable tool demanding our solicitousness. Recently, researchers have started to employ autonomous real-time systems involving deep learning techniques for this purpose. However, the wide va-riety of heterogeneous diseases affecting crop yield continues to prove itself a mammoth task for farmers and stymies the researchers. The majority of the current state-of-the-art models utilize datasets like Plant Village, consisting of leaf images taken in a controlled lab environment, which do not serve as accurate representative data of the real-world scenario. Moreover, the effectiveness of state-of-the-art models like EfficientNetLite, which provided notable improvements in accuracy for similar deep learning applications, remains untested on plant disease datasets. Hence, an exhaustive study on the performance of various state-of-the-art models with varied training conditions on real-time datasets like PlantDoc is imperative to further this area of research. In this paper, we have explored state-of-the-art CNNs like InceptionResNet, EfficientNetLite_0, and VGG-19, under various parameter settings, image augmentations techniques, and loss functions on the real-time PlantDoc dataset. We have evaluated and presented a comparative analysis of the exhaustive combinations and performances gauged by accuracy, top-5 accuracy, F1 scores, and other inferences drawn from training and testing all these networks on 27 different classes of crop diseases. We infer that EfficientNetLite proved to be a most effective architecture, especially given its relatively smaller size. EfficientNetLite coupled with the focal loss function and Albumentaions augmentation library yielded the best results with an accuracy of 70.71%, mean F1-score as 0.7, and 95.57% top-5 accuracy, on a test set congruent with real-world relevance.\n_____\n\n_____\n**[Ensemble Deep Learning Models for Fine-grained Plant Species Identification](https://doi.org/10.1109/CSDE53843.2021.9718387)**\n\n\n> **TL;DR:** This study investigates the development and application of ensemble deep learning models for fine-grained plant species identification. Six popular deep learning models (InceptionResNetV2, ResNet50, Xception, InceptionV3, MobileNetV2, and GoogleNet) were trained on both datasets and heterogeneous ensembles with various ensemble techniques (mean, weighted mean, voting, and stacked generalization) were performed. The validation and testing accuracy results for individual models were compared with the output generated by the ensemble methods.\n\nAutomated plant species identification for the datasets (images) collected from the natural environment is a challenging task. This study investigates the development and application of ensemble deep learning models for fine-grained plant species identification. Two different types of plant species datasets have been used in this study. The first dataset (UBD_45) consists of 45 medicinal plant species from the natural environment with the imbalanced distribution of classes and the second dataset (VP_200) has 200 medicinal plant species with balanced classes from the natural environment. Six popular deep learning models (InceptionResNetV2, ResNet50, Xception, InceptionV3, MobileNetV2, and GoogleNet) were trained on both datasets and heterogeneous ensembles with various ensemble techniques (mean, weighted mean, voting, and stacked generalization) were performed. The validation and testing accuracy results for individual models were compared with the output generated by the ensemble methods. The highest testing accuracies for base models were found 96.7% and 91.2% for UBD_45 and VP_200 datasets, respectively. Mean, weighted mean, and stacking ensembles showed better performance for both datasets. The stacking ensemble improved the classification accuracy by around 1.8% for the UBD_45 dataset while for VP_200 a significant improvement of around 4.23% was noticed using a weighted mean ensemble.\n_____\n\n_____\n**[A Deep Learning Approach for Plant Material Disease Identification](https://doi.org/10.1088/1757-899X/1116/1/012133)**\n\nPlant Material Disease Identification is essential for the food safety. To increase the crop production for the growing population of the world, the proper treatment is required on proper time to save the plant. Therefore, disease diagnosing on time is very important. This paper uses a deep learning convolutional neural network model to identify the plant disease. The pre-existing deep learning model Alexnet has been employed for plant disease identification in which an external feature of segmented plant material (leaves) is passed to the deepest fully connected layer. This combination of extracted feature by Alexnet and external feature of segmented plant material helps in plant disease identification. Experimental analysis has been done on a standard dataset Plant Village which has total 54,306 leaf images of 15 distinct plants having 38 diseases. The presented CNN approach worked well and outperformed to the existing approach.\n_____\n\n_____\n**[Plant Disease Identification Based on Deep Learning Algorithm in Smart Farming](https://doi.org/10.1155/2020/2479172)**\n\nThe identification of plant disease is the premise of the prevention of plant disease efficiently and precisely in the complex environment. With the rapid development of the smart farming, the identification of plant disease becomes digitalized and data-driven, enabling advanced decision support, smart analyses, and planning. This paper proposes a mathematical model of plant disease detection and recognition based on deep learning, which improves accuracy, generality, and training efficiency. Firstly, the region proposal network (RPN) is utilized to recognize and localize the leaves in complex surroundings. Then, images segmented based on the results of RPN algorithm contain the feature of symptoms through Chan–Vese (CV) algorithm. Finally, the segmented leaves are input into the transfer learning model and trained by the dataset of diseased leaves under simple background. Furthermore, the model is examined with black rot, bacterial plaque, and rust diseases. The results show that the accuracy of the method is 83.57%, which is better than the traditional method, thus reducing the influence of disease on agricultural production and being favorable to sustainable development of agriculture. Therefore, the deep learning algorithm proposed in the paper is of great significance in intelligent agriculture, ecological protection, and agricultural production.\n_____\n", "1778954": "Hi @satoshidatamoto! Thanks for sharing this important piece of information!" } }