{ "id": 323876, "title": " Read Papers About Machine Learning And Object Counting!", "url": "/competitions/iwildcam2022-fgvc9/discussion/323876", "author_name": "", "post_date": "2022-05-09T01:52:18", "votes": 2, "comment_count": 0, "views": 0, "content": "

'
\nThis is a list of papers that will make you smarter.
\nThis is a list of papers that will make you more thoughtful.
\nThis is a list of papers that will make you analyze the world in new ways.
\nThis is a list of papers that will make you see the world in a different light.
\nThis is a list of papers that might make you get a better score.
\nBecause this is what we are here for, right ;)
\nEnjoy! <3

\n
\n

An Advanced Deep Learning Approach for Multi-Object Counting in Urban Vehicular Environments
\nObject counting is an active research area that gained more attention in the past few years. In smart cities, vehicle counting plays a crucial role in urban planning and management of the Intelligent Transportation Systems (ITS). Several approaches have been proposed in the literature to address this problem. However, the resulting detection accuracy is still not adequate. This paper proposes an efficient approach that uses deep learning concepts and correlation filters for multi-object counting and tracking. The performance of the proposed system is evaluated using a dataset consisting of 16 videos with different features to examine the impact of object density, image quality, angle of view, and speed of motion towards system accuracy. Performance evaluation exhibits promising results in normal traffic scenarios and adverse weather conditions. Moreover, the proposed approach outperforms the performance of two recent approaches from the literature.

\n
\n
\n

Object Counting using Deep Learning
\nIn this paper, we consider the Problem of Object counting in deep learning. It is frequently carried out in different place of industries, school and colleges, traffic places among others. Object counting is major for quantitative analyses that rely on evaluation on certain objects. In this work, we propose a deep learning to find this challenge. Unfortunately, Object counting is most commonly a manual task and can be time intensive.  As a result, we manage both to increase the accuracy count and decrease the processing time. A Deep Learning based system can be used for real time applications.

\n
\n
\n

Towards Perspective-Free Object Counting with Deep Learning

\n
\n

TL;DR: Our models are able to precisely estimate the number of vehicles in a traffic congestion, or to count the humans in a very crowded scene. Our first contribution is the proposal of a novel convolutional neural network solution, named Counting CNN (CCNN). Essentially, the CCNN is formulated as a regression model where the network learns how to map the appearance of the image patches to their corresponding object density maps. Our second contribution consists in a scale-aware counting model, the Hydra CNN, able to estimate object densities in different very crowded scenarios where no geometric information of the scene can be provided. Hydra CNN learns a multisc
\n In this paper we address the problem of counting objects instances in images. Our models are able to precisely estimate the number of vehicles in a traffic congestion, or to count the humans in a very crowded scene. Our first contribution is the proposal of a novel convolutional neural network solution, named Counting CNN (CCNN). Essentially, the CCNN is formulated as a regression model where the network learns how to map the appearance of the image patches to their corresponding object density maps. Our second contribution consists in a scale-aware counting model, the Hydra CNN, able to estimate object densities in different very crowded scenarios where no geometric information of the scene can be provided. Hydra CNN learns a multiscale non-linear regression model which uses a pyramid of image patches extracted at multiple scales to perform the final density prediction. We report an extensive experimental evaluation, using up to three different object counting benchmarks, where we show how our solutions achieve a state-of-the-art performance.

\n
\n
\n

Everything counts: a Taxonomy of Deep Learning Approaches for Object Counting
\n TL;DR: We provide a taxonomy that reflects both: the method characteristic as well as the counting problem characteristics in order to derive a systematization of the scientific field. This taxonomy is applied on four case studies.
\n Many tasks, like process fault detection, disease diagnostic or maintaining security in public places are connected to the process of counting objects from observing image data like camera footage or microscopic images. However, the pivotal process of counting objects based on image data is subject to a lot of error sources when done manually. Therefore, a lot of approaches exist that aim to aid the automation of the counting process. Since most of those methods are of black box nature and have to be applied to real world scenarios, we provide a taxonomy that reflects both: the method characteristic as well as the counting problem characteristics in order to derive a systematization of the scientific field and to provide a tool that acts as a set of guidelines for choosing appropriate counting methods for different situations. We first conduct a literature review, where counting problem characteristics and solutions are extracted and later transformed into a taxonomy. We finally showcase the taxonomy using four case studies that are based on publicly available datasets for the sake of scientific comprehension.

\n
\n
\n

Crowdsourcing annotation system of object counting dataset for deep learning algorithm
\n Deep Learning is currently the state-of-the-art technique for various Computer Vision tasks, including object counting. Despite of its high performance, Deep Learning requires a gigantic amount of training data to show its best result. Getting this massive data in reasonable time requires a proper strategy such as crowdsourcing. However, in case of object counting, we found no crowdsourcing system able to effectively collect necessary data. To tackle this problem, we develop a crowdsourcing system to annotate image for object counting dataset. This system is also equipped with validation system to ensure the quality of collected dataset.

\n
\n
\n

Automatic Pest Counting from Pheromone Trap Images Using Deep Learning Object Detectors for Matsucoccus thunbergianae Monitoring
\n Simple Summary The black pine bast scale, Matsucoccus thunbergianae, is a forest pest that causes widespread damage to black pine; therefore, monitoring this pest is necessary to minimize environmental and economic losses in forests. However, monitoring insects in pheromone traps performed by humans is labor intensive and time consuming. To develop an automated monitoring system, we aimed to develop algorithms that detect and count M. thunbergianae from images of pheromone traps using deep-learning-based object detection algorithms. Object detection models based on deep learning neural networks under various conditions were trained, and the performances of detection and counting were compared and evaluated. In addition, the models were trained to detect small objects well by cropping images into multiple windows. As a result, the algorithms based on deep learning neural networks successfully detected and counted M. thunbergianae. These results showed that accurate and constant pest monitoring is possible using the artificial-intelligence-based methods we proposed. Abstract The black pine bast scale, M. thunbergianae, is a major insect pest of black pine and causes serious environmental and economic losses in forests. Therefore, it is essential to monitor the occurrence and population of M. thunbergianae, and a monitoring method using a pheromone trap is commonly employed. Because the counting of insects performed by humans in these pheromone traps is labor intensive and time consuming, this study proposes automated deep learning counting algorithms using pheromone trap images. The pheromone traps collected in the field were photographed in the laboratory, and the images were used for training, validation, and testing of the detection models. In addition, the image cropping method was applied for the successful detection of small objects in the image, considering the small size of M. thunbergianae in trap images. The detection and counting performance were evaluated and compared for a total of 16 models under eight model conditions and two cropping conditions, and a counting accuracy of 95% or more was shown in most models. This result shows that the artificial intelligence-based pest counting method proposed in this study is suitable for constant and accurate monitoring of insect pests.

\n
\n
\n

Counting Dense Leaves under Natural Environments via an Improved Deep-Learning-Based Object Detection Algorithm
\n The leaf is the organ that is crucial for photosynthesis and the production of nutrients in plants; as such, the number of leaves is one of the key indicators with which to describe the development and growth of a canopy. The irregular shape and distribution of the blades, as well as the effect of natural light, make the segmentation and detection process of the blades difficult. The inaccurate acquisition of plant phenotypic parameters may affect the subsequent judgment of crop growth status and crop yield. To address the challenge in counting dense and overlapped plant leaves under natural environments, we proposed an improved deep-learning-based object detection algorithm by merging a space-to-depth module, a Convolutional Block Attention Module (CBAM) and Atrous Spatial Pyramid Pooling (ASPP) into the network, and applying the smoothL1 function to improve the loss function of object prediction. We evaluated our method on images of five different plant species collected under indoor and outdoor environments. The experimental results demonstrated that our algorithm which counts dense leaves improved average detection accuracy of 85% to 96%. Our algorithm also showed better performance in both detection accuracy and time consumption compared to other state-of-the-art object detection algorithms.

\n
\n
\n

Plant Counting of Cotton from UAS Imagery Using Deep Learning-Based Object Detection Framework
\n Assessing plant population of cotton is important to make replanting decisions in low plant density areas, prone to yielding penalties. Since the measurement of plant population in the field is labor intensive and subject to error, in this study, a new approach of image-based plant counting is proposed, using unmanned aircraft systems (UAS; DJI Mavic 2 Pro, Shenzhen, China) data. The previously developed image-based techniques required a priori information of geometry or statistical characteristics of plant canopy features, while also limiting the versatility of the methods in variable field conditions. In this regard, a deep learning-based plant counting algorithm was proposed to reduce the number of input variables, and to remove requirements for acquiring geometric or statistical information. The object detection model named You Only Look Once version 3 (YOLOv3) and photogrammetry were utilized to separate, locate, and count cotton plants in the seedling stage. The proposed algorithm was tested with four different UAS datasets, containing variability in plant size, overall illumination, and background brightness. Root mean square error (RMSE) and R2 values of the optimal plant count results ranged from 0.50 to 0.60 plants per linear meter of row (number of plants within 1 m distance along the planting row direction) and 0.96 to 0.97, respectively. The object detection algorithm, trained with variable plant size, ground wetness, and lighting conditions generally resulted in a lower detection error, unless an observable difference of developmental stages of cotton existed. The proposed plant counting algorithm performed well with 0–14 plants per linear meter of row, when cotton plants are generally separable in the seedling stage. This study is expected to provide an automated methodology for in situ evaluation of plant emergence using UAS data.

\n
\n
\n

Vehicle Counting System using Deep Learning and Multi-Object Tracking Methods
\n Using deep learning technology and multi-object tracking method to count vehicles accurately in different traffic conditions is a hot research topic in the field of intelligent transportation. In this paper, first, a vehicle dataset from the perspective of highway surveillance cameras is constructed, and the vehicle detection model is obtained by training using the You Only Look Once (YOLO) version 3 network. Second, an improved multi-scale and multi-feature tracking algorithm based on a kernel correlation filter (KCF) algorithm is proposed to avoid the KCF extracting single features and single-scale defects. Combining the intersection over union (IoU) similarity measure and the row-column optimal association criterion proposed in this paper, matching strategy is used to process the vehicles that are not detected and wrongly detected, thereby obtaining complete vehicle trajectories. Finally, according to the trajectory of the vehicle, the traveling direction of the vehicle is automatically determined, and the setting position of the detecting line is automatically updated to obtain the vehicle count result accurately. Experiments were conducted in a variety of traffic scenes and compared with published data. The experimental results show that the proposed method achieves high accuracy of vehicle detection while maintaining accuracy and precision in tracking multiple objects, and obtains accurate vehicle counting results which can meet real-time processing requirements. The algorithm presented in this paper has practical application for vehicle counting in complex highway scenes.

\n
\n
\n

Occlusion Robust Wheat Ear Counting Algorithm Based on Deep Learning
\n Counting the number of wheat ears in images under natural light is an important way to evaluate the crop yield, thus, it is of great significance to modern intelligent agriculture. However, the distribution of wheat ears is dense, so the occlusion and overlap problem appears in almost every wheat image. It is difficult for traditional image processing methods to solve occlusion problem due to the deficiency of high-level semantic features, while existing deep learning based counting methods did not solve the occlusion efficiently. This article proposes an improved EfficientDet-D0 object detection model for wheat ear counting, and focuses on solving occlusion. First, the transfer learning method is employed in the pre-training of the model backbone network to extract the high-level semantic features of wheat ears. Secondly, an image augmentation method Random-Cutout is proposed, in which some rectangles are selected and erased according to the number and size of the wheat ears in the images to simulate occlusion in real wheat images. Finally, convolutional block attention module (CBAM) is adopted into the EfficientDet-D0 model after the backbone, which makes the model refine the features, pay more attention to the wheat ears and suppress other useless background information. Extensive experiments are done by feeding the features to detection layer, showing that the counting accuracy of the improved EfficientDet-D0 model reaches 94%, which is about 2% higher than the original model, and false detection rate is 5.8%, which is the lowest among comparative methods.

\n
\n
\n

Applications of object detection in modular construction based on a comparative evaluation of deep learning algorithms
\n Purpose
\n The practice of artificial intelligence (AI) is increasingly being promoted by technology developers. However, its adoption rate is still reported as low in the construction industry due to a lack of expertise and the limited reliable applications for AI technology. Hence, this paper aims to present the detailed outcome of experimentations evaluating the applicability and the performance of AI object detection algorithms for construction modular object detection.
\n Design/methodology/approach
\n This paper provides a thorough evaluation of two deep learning algorithms for object detection, including the faster region-based convolutional neural network (faster RCNN) and single shot multi-box detector (SSD). Two types of metrics are also presented; first, the average recall and mean average precision by image pixels; second, the recall and precision by counting. To conduct the experiments using the selected algorithms, four infrastructure and building construction sites are chosen to collect the required data, including a total of 990 images of three different but common modular objects, including modular panels, safety barricades and site fences.
\n Findings
\n The results of the comprehensive evaluation of the algorithms show that the performance of faster RCNN and SSD depends on the context that detection occurs. Indeed, surrounding objects and the backgrounds of the objects affect the level of accuracy obtained from the AI analysis and may particularly effect precision and recall. The analysis of loss lines shows that the loss lines for selected objects depend on both their geometry and the image background. The results on selected objects show that faster RCNN offers higher accuracy than SSD for detection of selected objects.
\n Research limitations/implications
\n The results show that modular object detection is crucial in construction for the achievement of the required information for project quality and safety objectives. The detection process can significantly improve monitoring object installation progress in an accurate and machine-based manner avoiding human errors. The results of this paper are limited to three construction sites, but future investigations can cover more tasks or objects from different construction sites in a fully automated manner.
\n Originality/value
\n This paper’s originality lies in offering new AI applications in modular construction, using a large first-hand data set collected from three construction sites. Furthermore, the paper presents the scientific evaluation results of implementing recent object detection algorithms across a set of extended metrics using the original training and validation data sets to improve the generalisability of the experimentation. This paper also provides the practitioners and scholars with a workflow on AI applications in the modular context and the first-hand referencing data.

\n
\n
\n

Leaf Counting in Rice (Oryza Sativa L.) Using Object Detection: A Deep Learning Approach
\n Leaf count is one of the crucial tasks in plant phenotyping, and leaves are the basic unit of plant architecture involved in photosynthesis, growth, and yield of a plant. Therefore, the total number of leaves per plant is considered as one of the essential physio-morphological plant traits for phenotyping. The current work proposes to estimate the total number of leaves of a rice plant by detecting their leaves tips. A rice plant has a single tip for a single leaf. Hence, this proposed framework counts the total number of leaves by counting the number of leaves tips equal to the number of leaves. You Only Look Once (YOLO) algorithm is used for the detection of the leaves tips as an object. This hypothesis builds a basis for counting the total number of leaves in a plant like rice, and similar field crops such as wheat (Triticum aestivum L), maize (Zea mays L.), sorghum (Sorghum bicolor), barley (Hordeum vulgare L.). The model detected leaves of a rice plant (RGB images) by detecting corresponding leaves tips with YOLO having average accuracy up to 82% and IOU around 0.53-0.60 and estimates the number of leaves in a plant by counting predicted bounding boxes around tips. The model also performed well with the wheat crop.

\n
\n
\n

Central Object Segmentation by Deep Learning to Continuously Monitor Fruit Growth through RGB Images
\n Monitoring fruit growth is useful when estimating final yields in advance and predicting optimum harvest times. However, observing fruit all day at the farm via RGB images is not an easy task because the light conditions are constantly changing. In this paper, we present CROP (Central Roundish Object Painter). The method involves image segmentation by deep learning, and the architecture of the neural network is a deeper version of U-Net. CROP identifies different types of central roundish fruit in an RGB image in varied light conditions, and creates a corresponding mask. Counting the mask pixels gives the relative two-dimensional size of the fruit, and in this way, time-series images may provide a non-contact means of automatically monitoring fruit growth. Although our measurement unit is different from the traditional one (length), we believe that shape identification potentially provides more information. Interestingly, CROP can have a more general use, working even for some other roundish objects. For this reason, we hope that CROP and our methodology yield big data to promote scientific advancements in horticultural science and other fields.

\n
\n
\n

A NOVEL DEEP LEARNING BASED METHOD FOR DETECTION AND COUNTING OF VEHICLES IN URBAN TRAFFIC SURVEILLANCE SYSTEMS
\n In intelligent transportation systems (ITS), it is essential to obtain reliable statistics of the vehicular flow in order to create urban traffic management strategies. These systems have benefited from the increase in computational resources and the improvement of image processing methods, especially in object detection based on deep learning. This paper proposes a method for vehicle counting composed of three stages: object detection, tracking and trajectory processing. In order to select the detection model with the best trade-off between accuracy and speed, the following one-stage detection models were compared: SSD512, CenterNet, Efficiedet-D0 and YOLO family models (v2, v3 and v4). Experimental results conducted on the benchmark dataset show that the best rates among the detection models were obtained using YOLOv4 with mAP =87% and a processing speed of 18 FPS. On the other hand, the accuracy obtained in the proposed counting method was 94% with a real-time processing rate lower than 1.9.

\n
\n
\n

YOLO-Based Deep Learning Framework for Olive Fruit Fly Detection and Counting
\n The olive fruit fly can damage up to 100% of the harvested fruit and can cause up to 80% reduction of the value of the resulting olive oil. Therefore, it is important to early detect its presence in the olive orchard to take the appropriate chemical or biological countermeasures as early as possible. Traps filled with attractant pheromones are typically deployed across the orchard to attract and capture the flies. Traditionally, the captured flies were manually counted which is error prone. Recently, the traps are employed with cameras and communication devices to send pictures of the captured flies to experts for analysis which is also error prone and inefficient. Consequently, machine and deep learning have been exploited to develop fully automated and accurate detection that does not include human in the loop. Such a learning problem is challenging due to the small size of the detected object, the differences in the light conditions at which pictures were taken, and the lack of enough data to train the learning model. In this paper, we present a deep learning framework for detecting and counting the number of olive fruit flies that exploits data augmentation to increase the dataset size, includes negative samples in the training to improve the detection accuracy, and normalizes the images to the color of the trap background, i.e., yellow, to unify the illumination conditions. The results of the proposed framework show a precision of 0.84, a recall of 0.97, an F1-score of 0.9 and mean Average Precision (mAP) of 96.68% which significantly outperforms existing pest detection systems.

\n
\n
\n

A Deep Learning Bidirectional Temporal Tracking Algorithm for Automated Blood Cell Counting from Non-invasive Capillaroscopy Videos
\n TL;DR: We propose a deep learning multi-cell tracking model, CycleTrack, which achieves accurate blood cell counting from capillaroscopic videos. Our approach combines two simple online tracking models, SORT and CenterTrack, and is tailored to features of capillary blood cell flow. Blood cells are tracked by displacement vectors in two opposing temporal directions (forward- and backward-tracking) between consecutive frames. The proposed model outperforms other baseline trackers, achieving 65.57% Multiple Object Tracking Accuracy and 73.95% ID F1 score on test videos. Compared to manual blood cell counting, CycleTrack achieves 96.58 $\\pm$
\n Oblique back-illumination capillaroscopy has recently been introduced as a method for high-quality, non-invasive blood cell imaging in human capillaries. To make this technique practical for clinical blood cell counting, solutions for automatic processing of acquired videos are needed. Here, we take the first step towards this goal, by introducing a deep learning multi-cell tracking model, named CycleTrack, which achieves accurate blood cell counting from capillaroscopic videos. CycleTrack combines two simple online tracking models, SORT and CenterTrack, and is tailored to features of capillary blood cell flow. Blood cells are tracked by displacement vectors in two opposing temporal directions (forward- and backward-tracking) between consecutive frames. This approach yields accurate tracking despite rapidly moving and deforming blood cells. The proposed model outperforms other baseline trackers, achieving 65.57% Multiple Object Tracking Accuracy and 73.95% ID F1 score on test videos. Compared to manual blood cell counting, CycleTrack achieves 96.58 $\\pm$ 2.43% cell counting accuracy among 8 test videos with 1000 frames each compared to 93.45% and 77.02% accuracy for independent CenterTrack and SORT almost without additional time expense. It takes 800s to track and count approximately 8000 blood cells from 9,600 frames captured in a typical one-minute video. Moreover, the blood cell velocity measured by CycleTrack demonstrates a consistent, pulsatile pattern within the physiological range of heart rate. Lastly, we discuss future improvements for the CycleTrack framework, which would enable clinical translation of the oblique back-illumination microscope towards a real-time and non-invasive point-of-care blood cell counting and analyzing technology.

\n
\n
\n

A NOVEL DEEP LEARNING BASED METHOD FOR DETECTION AND COUNTING OF VEHICLES IN URBAN TRAFFIC SURVEILLANCE SYSTEMS
\n TL;DR: This paper proposes a method for vehicle counting composed of three stages: object detection, tracking and trajectory processing. In order to select the detection model with the best trade-off between accuracy and speed, the following one-stage detection models were compared: SSD512, CenterNet, Efficiedet-D0 and YOLO family models (v2, v3 and v4). Experimental results conducted on the benchmark dataset show that the best rates among the detection models were obtained using YOLOv4 with mAP =87% and a processing speed of 18 FPS. On the other hand, the accuracy obtained in the
\n In intelligent transportation systems (ITS), it is essential to obtain reliable statistics of the vehicular flow in order to create urban traffic management strategies. These systems have benefited from the increase in computational resources and the improvement of image processing methods, especially in object detection based on deep learning. This paper proposes a method for vehicle counting composed of three stages: object detection, tracking and trajectory processing. In order to select the detection model with the best trade-off between accuracy and speed, the following one-stage detection models were compared: SSD512, CenterNet, Efficiedet-D0 and YOLO family models (v2, v3 and v4). Experimental results conducted on the benchmark dataset show that the best rates among the detection models were obtained using YOLOv4 with mAP =87% and a processing speed of 18 FPS. On the other hand, the accuracy obtained in the proposed counting method was 94% with a real-time processing rate lower than 1.9.

\n
\n
\n

Multiple Object Tracking using Deep Learning with YOLO V5
\n The MOT (Multiple Object Tracking) is an important tool in the modern world. It has various uses like object detection, counting objects, security tools ,etc. The Object tracking is a prominent technology in image processing which has a large future scope. The MOT has made significant growth in a few years due to deep learning, computer vision, machine learning, etc. This paper aims to provide a software solution that keeps track of the objects so that it can handle object list and count. By using YOLO “You Only Look Once” Technology with the help of Pytorch, the system aims in object detection, tracking and counting. Also unlike the general yolo object detection tool which detects all objects at the same time ,this MOT system also detects only objects which are needed to be detected by the user and thus helps in improving the performance of the system.

\n
\n
\n

Tomato Fruit Detection and Counting in Greenhouses Using Deep Learning
\n Accurately detecting and counting fruits during plant growth using imaging and computer vision is of importance not only from the point of view of reducing labor intensive manual measurements of phenotypic information, but also because it is a critical step toward automating processes such as harvesting. Deep learning based methods have emerged as the state-of-the-art techniques in many problems in image segmentation and classification, and have a lot of promise in challenging domains such as agriculture, where they can deal with the large variability in data better than classical computer vision methods. This paper reports results on the detection of tomatoes in images taken in a greenhouse, using the MaskRCNN algorithm, which detects objects and also the pixels corresponding to each object. Our experimental results on the detection of tomatoes from images taken in greenhouses using a RealSense camera are comparable to or better than the metrics reported by earlier work, even though those were obtained in laboratory conditions or using higher resolution images. Our results also show that MaskRCNN can implicitly learn object depth, which is necessary for background elimination.

\n
\n
\n

Towards a low-cost embedded vehicle counting system based on deep-learning for traffic management applications
\n TL;DR: The Kalman filter is more efficient than the centroid tracking algorithm, and it produces fewer errors.
\n This paper explores the feasibility of using a low-cost embedded system for real-time vehicle detection and counting through the use of deep neural networks. It compares the performance of two different object tracking methods, the Kalman filter with the Hungarian algorithm and the centroid tracking algorithm. The experimentation proved that the efficiency of the implemented algorithms was above the 92% and 98% for the centroid tracking algorithm and Kalman filter with the Hungarian algorithm, respectively. Also, the Kalman filter produced fewer errors overcoming the centroid tracking algorithm.

\n
\n
\n

Recognizing and counting Dendrocephalus brasiliensis (Crustacea: Anostraca) cysts using deep learning
\n TL;DR: We propose an automatized approach for the detection and counting of Dendrocephalus brasiliensis cysts from images captured by a digital microscope. For this purpose, we built the DBrasiliensis dataset, a repository with 246 images containing 5141 cysts of Dendrocephalus brasiliensis. Then, we trained two state-of-the-art object detection methods, YOLOv3 (You Only Look Once) and Faster R-CNN (Region-based Convolutional Neural Networks), on DBrasiliensis dataset in order to compare them under both cyst detection and counting tasks.
\n The Dendrocephalus brasiliensis, a native species from South America, is a freshwater crustacean well explored in conservational and productive activities. Its main characteristics are its rusticity and resistance cysts production, in which the hatching requires a period of dehydration. Independent of the species utilization nature, it is essential to manipulate its cysts, such as the counting using microscopes. Manually counting is a difficult task, prone to errors, and that also very time-consuming. In this paper, we propose an automatized approach for the detection and counting of Dendrocephalus brasiliensis cysts from images captured by a digital microscope. For this purpose, we built the DBrasiliensis dataset, a repository with 246 images containing 5141 cysts of Dendrocephalus brasiliensis. Then, we trained two state-of-the-art object detection methods, YOLOv3 (You Only Look Once) and Faster R-CNN (Region-based Convolutional Neural Networks), on DBrasiliensis dataset in order to compare them under both cyst detection and counting tasks. Experiments showed evidence that YOLOv3 is superior to Faster R-CNN, achieving an accuracy rate of 83,74%, R2 of 0.88, RMSE (Root Mean Square Error) of 3.49, and MAE (Mean Absolute Error) of 2.24 on cyst detection and counting. Moreover, we showed that is possible to infer the number of cysts of a substrate, with known weight, by performing the automated counting of some of its samples. In conclusion, the proposed approach using YOLOv3 is adequate to detect and count Dendrocephalus brasiliensis cysts. The DBrasiliensis dataset can be accessed at: https://doi.org/10.6084/m9.figshare.13073240.

\n
\n
\n

SpikeSegNet-a deep learning approach utilizing encoder-decoder network with hourglass for spike segmentation and counting in wheat plant from visual imaging
\n Background High throughput non-destructive phenotyping is emerging as a significant approach for phenotyping germplasm and breeding populations for the identification of superior donors, elite lines, and QTLs. Detection and counting of spikes, the grain bearing organs of wheat, is critical for phenomics of a large set of germplasm and breeding lines in controlled and field conditions. It is also required for precision agriculture where the application of nitrogen, water, and other inputs at this critical stage is necessary. Further, counting of spikes is an important measure to determine yield. Digital image analysis and machine learning techniques play an essential role in non-destructive plant phenotyping analysis. Results In this study, an approach based on computer vision, particularly object detection, to recognize and count the number of spikes of the wheat plant from the digital images is proposed. For spike identification, a novel deep-learning network, SpikeSegNet, has been developed by combining two proposed feature networks: Local Patch extraction Network (LPNet) and Global Mask refinement Network (GMRNet). In LPNet, the contextual and spatial features are learned at the local patch level. The output of LPNet is a segmented mask image, which is further refined at the global level using GMRNet. Visual (RGB) images of 200 wheat plants were captured using LemnaTec imaging system installed at Nanaji Deshmukh Plant Phenomics Centre, ICAR-IARI, New Delhi. The precision, accuracy, and robustness (F 1 score) of the proposed approach for spike segmentation are found to be 99.93%, 99.91%, and 99.91%, respectively. For counting the number of spikes, “analyse particles”—function of imageJ was applied on the output image of the proposed SpikeSegNet model. For spike counting, the average precision, accuracy, and robustness are 99%, 95%, and 97%, respectively. SpikeSegNet approach is tested for robustness with illuminated image dataset, and no significant difference is observed in the segmentation performance. Conclusion In this study, a new approach called as SpikeSegNet has been proposed based on combined digital image analysis and deep learning techniques. A dedicated deep learning approach has been developed to identify and count spikes in the wheat plants. The performance of the approach demonstrates that SpikeSegNet is an effective and robust approach for spike detection and counting. As detection and counting of wheat spikes are closely related to the crop yield, and the proposed approach is also non-destructive, it is a significant step forward in the area of non-destructive and high-throughput phenotyping of wheat.

\n
\n
\n

A Simple Vehicle Counting System Using Deep Learning with YOLOv3 Model
\n Deep Learning is a popular Machine Learning algorithm that is widely used in many areas in current daily life. Its robust performance and ready-to-use frameworks and architectures enables many people to develop various Deep Learning-based software or systems to support human tasks and activities. Traffic monitoring is one area that utilizes Deep Learning for several purposes. By using cameras installed in some spots on the roads, many tasks such as vehicle counting, vehicle identification, traffic violation monitoring, vehicle speed monitoring, etc. can be realized. In this paper, we discuss a Deep Learning implementation to create a vehicle counting system without having to track the vehicles movements. To enhance the system performance and to reduce time in deploying Deep Learning architecture, hence pretrained model of YOLOv3 is used in this research due to its good performance and moderate computational time in object detection. This research aims to create a simple vehicle counting system to help human in classify and counting the vehicles that cross the street. The counting is based on four types of vehicle, i.e. car, motorcycle, bus, and truck, while previous research counts the car only. As the result, our proposed system capable to count the vehicles crossing the road based on video captured by camera with the highest accuracy of 97.72%.

\n
\n
\n

Accurate stacked-sheet counting method based on deep learning.
\n TL;DR: This paper proposes a new stacked-sheet counting method with a deep learning approach using the U-Net architecture. A specific dataset according to the characteristics of stack side images is collected. The stripe of the center line of each sheet is used for semantic segmentation, and the complete side images of the slices are segmented via training with small image patches and testing with original large images. With this model, each pixel is classified by multi-layer convolution and deconvolution to determine whether it is the target object to be detected. After the model is trained, the test set is used to test the model, and a center region
\n The accurate counting of laminated sheets, such as packing or printing sheets in industry, is extremely important because it greatly affects the economic cost. However, the different thicknesses, adhesion properties, and breakage points and the low contrast of sheets remain challenges to traditional counting methods based on image processing. This paper proposes a new stacked-sheet counting method with a deep learning approach using the U-Net architecture. A specific dataset according to the characteristics of stack side images is collected. The stripe of the center line of each sheet is used for semantic segmentation, and the complete side images of the slices are segmented via training with small image patches and testing with original large images. With this model, each pixel is classified by multi-layer convolution and deconvolution to determine whether it is the target object to be detected. After the model is trained, the test set is used to test the model, and a center region segmentation map based on the pixel points is obtained. By calculating the statistical median value of centerline points across different sections in these segmented images, the number of sheets can be obtained. Compared with traditional image algorithms in real product counting experiments, the proposed method can achieve better performance with higher accuracy and a lower error rate.

\n
\n
\n

Automatic Detection and Counting of Lymphocytes from Immunohistochemistry Cancer Images Using Deep Learning
\n Purpose Cancer is one of the most life-threatening and devastating diseases in the world. The generally recognized standard for cancer staging is the TNM staging system. In addition, a new parameter called the Immunoscore has been developed recently to assess the survival rate of patients. The Immunoscore is based on counts of CD3+ and CD8+ lymphocytes in the tumour core and margin. Counting the number of lymphocytes is a tedious process for pathologists. This paper examines the use of deep learning techniques for automatic detection and counting of lymphocytes from immunohistochemistry images of breast, colon, and prostate cancers. Methods We used an object detector called Faster R-CNN with four feature extractors: Resnet-50, VGG-16, Inception-V2, and Resnet-101 for automatic detection and counting of lymphocytes. A total of 11,136 lymphocytes were annotated after performing data augmentation on 1228 images. The test images are separated into three regions of interest (ROI): scattered lymphocytes, groups of lymphocytes, and artefacts. In each ROI, the performance of the object detector was checked by evaluation metrics. Results On comparing the F1-score for all three ROIs, we found that Resnet-101 provides better performance than the other feature extractors when using Faster R-CNN. The mean error in lymphocyte count for all ROIs appears to be insignificant. The detection time for a single image is less for VGG-16 than for other feature extractors. Conclusion This study presents a fine-tuned Faster R-CNN object detector for automatic detection and counting of lymphocytes in three different cancer tissues for scoring lymphocytes. Our results suggest that the Faster R-CNN method is efficient and yields good results. Thus, the proposed method can assist pathologists in providing a cancer prognosis.

\n
\n
", "messages": [ { "id": 1781810, "postDate": "2022-05-09T01:52:18Z", "content": "

'
\nThis is a list of papers that will make you smarter.
\nThis is a list of papers that will make you more thoughtful.
\nThis is a list of papers that will make you analyze the world in new ways.
\nThis is a list of papers that will make you see the world in a different light.
\nThis is a list of papers that might make you get a better score.
\nBecause this is what we are here for, right ;)
\nEnjoy! <3

\n
\n

An Advanced Deep Learning Approach for Multi-Object Counting in Urban Vehicular Environments
\nObject counting is an active research area that gained more attention in the past few years. In smart cities, vehicle counting plays a crucial role in urban planning and management of the Intelligent Transportation Systems (ITS). Several approaches have been proposed in the literature to address this problem. However, the resulting detection accuracy is still not adequate. This paper proposes an efficient approach that uses deep learning concepts and correlation filters for multi-object counting and tracking. The performance of the proposed system is evaluated using a dataset consisting of 16 videos with different features to examine the impact of object density, image quality, angle of view, and speed of motion towards system accuracy. Performance evaluation exhibits promising results in normal traffic scenarios and adverse weather conditions. Moreover, the proposed approach outperforms the performance of two recent approaches from the literature.

\n
\n
\n

Object Counting using Deep Learning
\nIn this paper, we consider the Problem of Object counting in deep learning. It is frequently carried out in different place of industries, school and colleges, traffic places among others. Object counting is major for quantitative analyses that rely on evaluation on certain objects. In this work, we propose a deep learning to find this challenge. Unfortunately, Object counting is most commonly a manual task and can be time intensive.  As a result, we manage both to increase the accuracy count and decrease the processing time. A Deep Learning based system can be used for real time applications.

\n
\n
\n

Towards Perspective-Free Object Counting with Deep Learning

\n
\n

TL;DR: Our models are able to precisely estimate the number of vehicles in a traffic congestion, or to count the humans in a very crowded scene. Our first contribution is the proposal of a novel convolutional neural network solution, named Counting CNN (CCNN). Essentially, the CCNN is formulated as a regression model where the network learns how to map the appearance of the image patches to their corresponding object density maps. Our second contribution consists in a scale-aware counting model, the Hydra CNN, able to estimate object densities in different very crowded scenarios where no geometric information of the scene can be provided. Hydra CNN learns a multisc
\n In this paper we address the problem of counting objects instances in images. Our models are able to precisely estimate the number of vehicles in a traffic congestion, or to count the humans in a very crowded scene. Our first contribution is the proposal of a novel convolutional neural network solution, named Counting CNN (CCNN). Essentially, the CCNN is formulated as a regression model where the network learns how to map the appearance of the image patches to their corresponding object density maps. Our second contribution consists in a scale-aware counting model, the Hydra CNN, able to estimate object densities in different very crowded scenarios where no geometric information of the scene can be provided. Hydra CNN learns a multiscale non-linear regression model which uses a pyramid of image patches extracted at multiple scales to perform the final density prediction. We report an extensive experimental evaluation, using up to three different object counting benchmarks, where we show how our solutions achieve a state-of-the-art performance.

\n
\n
\n

Everything counts: a Taxonomy of Deep Learning Approaches for Object Counting
\n TL;DR: We provide a taxonomy that reflects both: the method characteristic as well as the counting problem characteristics in order to derive a systematization of the scientific field. This taxonomy is applied on four case studies.
\n Many tasks, like process fault detection, disease diagnostic or maintaining security in public places are connected to the process of counting objects from observing image data like camera footage or microscopic images. However, the pivotal process of counting objects based on image data is subject to a lot of error sources when done manually. Therefore, a lot of approaches exist that aim to aid the automation of the counting process. Since most of those methods are of black box nature and have to be applied to real world scenarios, we provide a taxonomy that reflects both: the method characteristic as well as the counting problem characteristics in order to derive a systematization of the scientific field and to provide a tool that acts as a set of guidelines for choosing appropriate counting methods for different situations. We first conduct a literature review, where counting problem characteristics and solutions are extracted and later transformed into a taxonomy. We finally showcase the taxonomy using four case studies that are based on publicly available datasets for the sake of scientific comprehension.

\n
\n
\n

Crowdsourcing annotation system of object counting dataset for deep learning algorithm
\n Deep Learning is currently the state-of-the-art technique for various Computer Vision tasks, including object counting. Despite of its high performance, Deep Learning requires a gigantic amount of training data to show its best result. Getting this massive data in reasonable time requires a proper strategy such as crowdsourcing. However, in case of object counting, we found no crowdsourcing system able to effectively collect necessary data. To tackle this problem, we develop a crowdsourcing system to annotate image for object counting dataset. This system is also equipped with validation system to ensure the quality of collected dataset.

\n
\n
\n

Automatic Pest Counting from Pheromone Trap Images Using Deep Learning Object Detectors for Matsucoccus thunbergianae Monitoring
\n Simple Summary The black pine bast scale, Matsucoccus thunbergianae, is a forest pest that causes widespread damage to black pine; therefore, monitoring this pest is necessary to minimize environmental and economic losses in forests. However, monitoring insects in pheromone traps performed by humans is labor intensive and time consuming. To develop an automated monitoring system, we aimed to develop algorithms that detect and count M. thunbergianae from images of pheromone traps using deep-learning-based object detection algorithms. Object detection models based on deep learning neural networks under various conditions were trained, and the performances of detection and counting were compared and evaluated. In addition, the models were trained to detect small objects well by cropping images into multiple windows. As a result, the algorithms based on deep learning neural networks successfully detected and counted M. thunbergianae. These results showed that accurate and constant pest monitoring is possible using the artificial-intelligence-based methods we proposed. Abstract The black pine bast scale, M. thunbergianae, is a major insect pest of black pine and causes serious environmental and economic losses in forests. Therefore, it is essential to monitor the occurrence and population of M. thunbergianae, and a monitoring method using a pheromone trap is commonly employed. Because the counting of insects performed by humans in these pheromone traps is labor intensive and time consuming, this study proposes automated deep learning counting algorithms using pheromone trap images. The pheromone traps collected in the field were photographed in the laboratory, and the images were used for training, validation, and testing of the detection models. In addition, the image cropping method was applied for the successful detection of small objects in the image, considering the small size of M. thunbergianae in trap images. The detection and counting performance were evaluated and compared for a total of 16 models under eight model conditions and two cropping conditions, and a counting accuracy of 95% or more was shown in most models. This result shows that the artificial intelligence-based pest counting method proposed in this study is suitable for constant and accurate monitoring of insect pests.

\n
\n
\n

Counting Dense Leaves under Natural Environments via an Improved Deep-Learning-Based Object Detection Algorithm
\n The leaf is the organ that is crucial for photosynthesis and the production of nutrients in plants; as such, the number of leaves is one of the key indicators with which to describe the development and growth of a canopy. The irregular shape and distribution of the blades, as well as the effect of natural light, make the segmentation and detection process of the blades difficult. The inaccurate acquisition of plant phenotypic parameters may affect the subsequent judgment of crop growth status and crop yield. To address the challenge in counting dense and overlapped plant leaves under natural environments, we proposed an improved deep-learning-based object detection algorithm by merging a space-to-depth module, a Convolutional Block Attention Module (CBAM) and Atrous Spatial Pyramid Pooling (ASPP) into the network, and applying the smoothL1 function to improve the loss function of object prediction. We evaluated our method on images of five different plant species collected under indoor and outdoor environments. The experimental results demonstrated that our algorithm which counts dense leaves improved average detection accuracy of 85% to 96%. Our algorithm also showed better performance in both detection accuracy and time consumption compared to other state-of-the-art object detection algorithms.

\n
\n
\n

Plant Counting of Cotton from UAS Imagery Using Deep Learning-Based Object Detection Framework
\n Assessing plant population of cotton is important to make replanting decisions in low plant density areas, prone to yielding penalties. Since the measurement of plant population in the field is labor intensive and subject to error, in this study, a new approach of image-based plant counting is proposed, using unmanned aircraft systems (UAS; DJI Mavic 2 Pro, Shenzhen, China) data. The previously developed image-based techniques required a priori information of geometry or statistical characteristics of plant canopy features, while also limiting the versatility of the methods in variable field conditions. In this regard, a deep learning-based plant counting algorithm was proposed to reduce the number of input variables, and to remove requirements for acquiring geometric or statistical information. The object detection model named You Only Look Once version 3 (YOLOv3) and photogrammetry were utilized to separate, locate, and count cotton plants in the seedling stage. The proposed algorithm was tested with four different UAS datasets, containing variability in plant size, overall illumination, and background brightness. Root mean square error (RMSE) and R2 values of the optimal plant count results ranged from 0.50 to 0.60 plants per linear meter of row (number of plants within 1 m distance along the planting row direction) and 0.96 to 0.97, respectively. The object detection algorithm, trained with variable plant size, ground wetness, and lighting conditions generally resulted in a lower detection error, unless an observable difference of developmental stages of cotton existed. The proposed plant counting algorithm performed well with 0–14 plants per linear meter of row, when cotton plants are generally separable in the seedling stage. This study is expected to provide an automated methodology for in situ evaluation of plant emergence using UAS data.

\n
\n
\n

Vehicle Counting System using Deep Learning and Multi-Object Tracking Methods
\n Using deep learning technology and multi-object tracking method to count vehicles accurately in different traffic conditions is a hot research topic in the field of intelligent transportation. In this paper, first, a vehicle dataset from the perspective of highway surveillance cameras is constructed, and the vehicle detection model is obtained by training using the You Only Look Once (YOLO) version 3 network. Second, an improved multi-scale and multi-feature tracking algorithm based on a kernel correlation filter (KCF) algorithm is proposed to avoid the KCF extracting single features and single-scale defects. Combining the intersection over union (IoU) similarity measure and the row-column optimal association criterion proposed in this paper, matching strategy is used to process the vehicles that are not detected and wrongly detected, thereby obtaining complete vehicle trajectories. Finally, according to the trajectory of the vehicle, the traveling direction of the vehicle is automatically determined, and the setting position of the detecting line is automatically updated to obtain the vehicle count result accurately. Experiments were conducted in a variety of traffic scenes and compared with published data. The experimental results show that the proposed method achieves high accuracy of vehicle detection while maintaining accuracy and precision in tracking multiple objects, and obtains accurate vehicle counting results which can meet real-time processing requirements. The algorithm presented in this paper has practical application for vehicle counting in complex highway scenes.

\n
\n
\n

Occlusion Robust Wheat Ear Counting Algorithm Based on Deep Learning
\n Counting the number of wheat ears in images under natural light is an important way to evaluate the crop yield, thus, it is of great significance to modern intelligent agriculture. However, the distribution of wheat ears is dense, so the occlusion and overlap problem appears in almost every wheat image. It is difficult for traditional image processing methods to solve occlusion problem due to the deficiency of high-level semantic features, while existing deep learning based counting methods did not solve the occlusion efficiently. This article proposes an improved EfficientDet-D0 object detection model for wheat ear counting, and focuses on solving occlusion. First, the transfer learning method is employed in the pre-training of the model backbone network to extract the high-level semantic features of wheat ears. Secondly, an image augmentation method Random-Cutout is proposed, in which some rectangles are selected and erased according to the number and size of the wheat ears in the images to simulate occlusion in real wheat images. Finally, convolutional block attention module (CBAM) is adopted into the EfficientDet-D0 model after the backbone, which makes the model refine the features, pay more attention to the wheat ears and suppress other useless background information. Extensive experiments are done by feeding the features to detection layer, showing that the counting accuracy of the improved EfficientDet-D0 model reaches 94%, which is about 2% higher than the original model, and false detection rate is 5.8%, which is the lowest among comparative methods.

\n
\n
\n

Applications of object detection in modular construction based on a comparative evaluation of deep learning algorithms
\n Purpose
\n The practice of artificial intelligence (AI) is increasingly being promoted by technology developers. However, its adoption rate is still reported as low in the construction industry due to a lack of expertise and the limited reliable applications for AI technology. Hence, this paper aims to present the detailed outcome of experimentations evaluating the applicability and the performance of AI object detection algorithms for construction modular object detection.
\n Design/methodology/approach
\n This paper provides a thorough evaluation of two deep learning algorithms for object detection, including the faster region-based convolutional neural network (faster RCNN) and single shot multi-box detector (SSD). Two types of metrics are also presented; first, the average recall and mean average precision by image pixels; second, the recall and precision by counting. To conduct the experiments using the selected algorithms, four infrastructure and building construction sites are chosen to collect the required data, including a total of 990 images of three different but common modular objects, including modular panels, safety barricades and site fences.
\n Findings
\n The results of the comprehensive evaluation of the algorithms show that the performance of faster RCNN and SSD depends on the context that detection occurs. Indeed, surrounding objects and the backgrounds of the objects affect the level of accuracy obtained from the AI analysis and may particularly effect precision and recall. The analysis of loss lines shows that the loss lines for selected objects depend on both their geometry and the image background. The results on selected objects show that faster RCNN offers higher accuracy than SSD for detection of selected objects.
\n Research limitations/implications
\n The results show that modular object detection is crucial in construction for the achievement of the required information for project quality and safety objectives. The detection process can significantly improve monitoring object installation progress in an accurate and machine-based manner avoiding human errors. The results of this paper are limited to three construction sites, but future investigations can cover more tasks or objects from different construction sites in a fully automated manner.
\n Originality/value
\n This paper’s originality lies in offering new AI applications in modular construction, using a large first-hand data set collected from three construction sites. Furthermore, the paper presents the scientific evaluation results of implementing recent object detection algorithms across a set of extended metrics using the original training and validation data sets to improve the generalisability of the experimentation. This paper also provides the practitioners and scholars with a workflow on AI applications in the modular context and the first-hand referencing data.

\n
\n
\n

Leaf Counting in Rice (Oryza Sativa L.) Using Object Detection: A Deep Learning Approach
\n Leaf count is one of the crucial tasks in plant phenotyping, and leaves are the basic unit of plant architecture involved in photosynthesis, growth, and yield of a plant. Therefore, the total number of leaves per plant is considered as one of the essential physio-morphological plant traits for phenotyping. The current work proposes to estimate the total number of leaves of a rice plant by detecting their leaves tips. A rice plant has a single tip for a single leaf. Hence, this proposed framework counts the total number of leaves by counting the number of leaves tips equal to the number of leaves. You Only Look Once (YOLO) algorithm is used for the detection of the leaves tips as an object. This hypothesis builds a basis for counting the total number of leaves in a plant like rice, and similar field crops such as wheat (Triticum aestivum L), maize (Zea mays L.), sorghum (Sorghum bicolor), barley (Hordeum vulgare L.). The model detected leaves of a rice plant (RGB images) by detecting corresponding leaves tips with YOLO having average accuracy up to 82% and IOU around 0.53-0.60 and estimates the number of leaves in a plant by counting predicted bounding boxes around tips. The model also performed well with the wheat crop.

\n
\n
\n

Central Object Segmentation by Deep Learning to Continuously Monitor Fruit Growth through RGB Images
\n Monitoring fruit growth is useful when estimating final yields in advance and predicting optimum harvest times. However, observing fruit all day at the farm via RGB images is not an easy task because the light conditions are constantly changing. In this paper, we present CROP (Central Roundish Object Painter). The method involves image segmentation by deep learning, and the architecture of the neural network is a deeper version of U-Net. CROP identifies different types of central roundish fruit in an RGB image in varied light conditions, and creates a corresponding mask. Counting the mask pixels gives the relative two-dimensional size of the fruit, and in this way, time-series images may provide a non-contact means of automatically monitoring fruit growth. Although our measurement unit is different from the traditional one (length), we believe that shape identification potentially provides more information. Interestingly, CROP can have a more general use, working even for some other roundish objects. For this reason, we hope that CROP and our methodology yield big data to promote scientific advancements in horticultural science and other fields.

\n
\n
\n

A NOVEL DEEP LEARNING BASED METHOD FOR DETECTION AND COUNTING OF VEHICLES IN URBAN TRAFFIC SURVEILLANCE SYSTEMS
\n In intelligent transportation systems (ITS), it is essential to obtain reliable statistics of the vehicular flow in order to create urban traffic management strategies. These systems have benefited from the increase in computational resources and the improvement of image processing methods, especially in object detection based on deep learning. This paper proposes a method for vehicle counting composed of three stages: object detection, tracking and trajectory processing. In order to select the detection model with the best trade-off between accuracy and speed, the following one-stage detection models were compared: SSD512, CenterNet, Efficiedet-D0 and YOLO family models (v2, v3 and v4). Experimental results conducted on the benchmark dataset show that the best rates among the detection models were obtained using YOLOv4 with mAP =87% and a processing speed of 18 FPS. On the other hand, the accuracy obtained in the proposed counting method was 94% with a real-time processing rate lower than 1.9.

\n
\n
\n

YOLO-Based Deep Learning Framework for Olive Fruit Fly Detection and Counting
\n The olive fruit fly can damage up to 100% of the harvested fruit and can cause up to 80% reduction of the value of the resulting olive oil. Therefore, it is important to early detect its presence in the olive orchard to take the appropriate chemical or biological countermeasures as early as possible. Traps filled with attractant pheromones are typically deployed across the orchard to attract and capture the flies. Traditionally, the captured flies were manually counted which is error prone. Recently, the traps are employed with cameras and communication devices to send pictures of the captured flies to experts for analysis which is also error prone and inefficient. Consequently, machine and deep learning have been exploited to develop fully automated and accurate detection that does not include human in the loop. Such a learning problem is challenging due to the small size of the detected object, the differences in the light conditions at which pictures were taken, and the lack of enough data to train the learning model. In this paper, we present a deep learning framework for detecting and counting the number of olive fruit flies that exploits data augmentation to increase the dataset size, includes negative samples in the training to improve the detection accuracy, and normalizes the images to the color of the trap background, i.e., yellow, to unify the illumination conditions. The results of the proposed framework show a precision of 0.84, a recall of 0.97, an F1-score of 0.9 and mean Average Precision (mAP) of 96.68% which significantly outperforms existing pest detection systems.

\n
\n
\n

A Deep Learning Bidirectional Temporal Tracking Algorithm for Automated Blood Cell Counting from Non-invasive Capillaroscopy Videos
\n TL;DR: We propose a deep learning multi-cell tracking model, CycleTrack, which achieves accurate blood cell counting from capillaroscopic videos. Our approach combines two simple online tracking models, SORT and CenterTrack, and is tailored to features of capillary blood cell flow. Blood cells are tracked by displacement vectors in two opposing temporal directions (forward- and backward-tracking) between consecutive frames. The proposed model outperforms other baseline trackers, achieving 65.57% Multiple Object Tracking Accuracy and 73.95% ID F1 score on test videos. Compared to manual blood cell counting, CycleTrack achieves 96.58 $\\pm$
\n Oblique back-illumination capillaroscopy has recently been introduced as a method for high-quality, non-invasive blood cell imaging in human capillaries. To make this technique practical for clinical blood cell counting, solutions for automatic processing of acquired videos are needed. Here, we take the first step towards this goal, by introducing a deep learning multi-cell tracking model, named CycleTrack, which achieves accurate blood cell counting from capillaroscopic videos. CycleTrack combines two simple online tracking models, SORT and CenterTrack, and is tailored to features of capillary blood cell flow. Blood cells are tracked by displacement vectors in two opposing temporal directions (forward- and backward-tracking) between consecutive frames. This approach yields accurate tracking despite rapidly moving and deforming blood cells. The proposed model outperforms other baseline trackers, achieving 65.57% Multiple Object Tracking Accuracy and 73.95% ID F1 score on test videos. Compared to manual blood cell counting, CycleTrack achieves 96.58 $\\pm$ 2.43% cell counting accuracy among 8 test videos with 1000 frames each compared to 93.45% and 77.02% accuracy for independent CenterTrack and SORT almost without additional time expense. It takes 800s to track and count approximately 8000 blood cells from 9,600 frames captured in a typical one-minute video. Moreover, the blood cell velocity measured by CycleTrack demonstrates a consistent, pulsatile pattern within the physiological range of heart rate. Lastly, we discuss future improvements for the CycleTrack framework, which would enable clinical translation of the oblique back-illumination microscope towards a real-time and non-invasive point-of-care blood cell counting and analyzing technology.

\n
\n
\n

A NOVEL DEEP LEARNING BASED METHOD FOR DETECTION AND COUNTING OF VEHICLES IN URBAN TRAFFIC SURVEILLANCE SYSTEMS
\n TL;DR: This paper proposes a method for vehicle counting composed of three stages: object detection, tracking and trajectory processing. In order to select the detection model with the best trade-off between accuracy and speed, the following one-stage detection models were compared: SSD512, CenterNet, Efficiedet-D0 and YOLO family models (v2, v3 and v4). Experimental results conducted on the benchmark dataset show that the best rates among the detection models were obtained using YOLOv4 with mAP =87% and a processing speed of 18 FPS. On the other hand, the accuracy obtained in the
\n In intelligent transportation systems (ITS), it is essential to obtain reliable statistics of the vehicular flow in order to create urban traffic management strategies. These systems have benefited from the increase in computational resources and the improvement of image processing methods, especially in object detection based on deep learning. This paper proposes a method for vehicle counting composed of three stages: object detection, tracking and trajectory processing. In order to select the detection model with the best trade-off between accuracy and speed, the following one-stage detection models were compared: SSD512, CenterNet, Efficiedet-D0 and YOLO family models (v2, v3 and v4). Experimental results conducted on the benchmark dataset show that the best rates among the detection models were obtained using YOLOv4 with mAP =87% and a processing speed of 18 FPS. On the other hand, the accuracy obtained in the proposed counting method was 94% with a real-time processing rate lower than 1.9.

\n
\n
\n

Multiple Object Tracking using Deep Learning with YOLO V5
\n The MOT (Multiple Object Tracking) is an important tool in the modern world. It has various uses like object detection, counting objects, security tools ,etc. The Object tracking is a prominent technology in image processing which has a large future scope. The MOT has made significant growth in a few years due to deep learning, computer vision, machine learning, etc. This paper aims to provide a software solution that keeps track of the objects so that it can handle object list and count. By using YOLO “You Only Look Once” Technology with the help of Pytorch, the system aims in object detection, tracking and counting. Also unlike the general yolo object detection tool which detects all objects at the same time ,this MOT system also detects only objects which are needed to be detected by the user and thus helps in improving the performance of the system.

\n
\n
\n

Tomato Fruit Detection and Counting in Greenhouses Using Deep Learning
\n Accurately detecting and counting fruits during plant growth using imaging and computer vision is of importance not only from the point of view of reducing labor intensive manual measurements of phenotypic information, but also because it is a critical step toward automating processes such as harvesting. Deep learning based methods have emerged as the state-of-the-art techniques in many problems in image segmentation and classification, and have a lot of promise in challenging domains such as agriculture, where they can deal with the large variability in data better than classical computer vision methods. This paper reports results on the detection of tomatoes in images taken in a greenhouse, using the MaskRCNN algorithm, which detects objects and also the pixels corresponding to each object. Our experimental results on the detection of tomatoes from images taken in greenhouses using a RealSense camera are comparable to or better than the metrics reported by earlier work, even though those were obtained in laboratory conditions or using higher resolution images. Our results also show that MaskRCNN can implicitly learn object depth, which is necessary for background elimination.

\n
\n
\n

Towards a low-cost embedded vehicle counting system based on deep-learning for traffic management applications
\n TL;DR: The Kalman filter is more efficient than the centroid tracking algorithm, and it produces fewer errors.
\n This paper explores the feasibility of using a low-cost embedded system for real-time vehicle detection and counting through the use of deep neural networks. It compares the performance of two different object tracking methods, the Kalman filter with the Hungarian algorithm and the centroid tracking algorithm. The experimentation proved that the efficiency of the implemented algorithms was above the 92% and 98% for the centroid tracking algorithm and Kalman filter with the Hungarian algorithm, respectively. Also, the Kalman filter produced fewer errors overcoming the centroid tracking algorithm.

\n
\n
\n

Recognizing and counting Dendrocephalus brasiliensis (Crustacea: Anostraca) cysts using deep learning
\n TL;DR: We propose an automatized approach for the detection and counting of Dendrocephalus brasiliensis cysts from images captured by a digital microscope. For this purpose, we built the DBrasiliensis dataset, a repository with 246 images containing 5141 cysts of Dendrocephalus brasiliensis. Then, we trained two state-of-the-art object detection methods, YOLOv3 (You Only Look Once) and Faster R-CNN (Region-based Convolutional Neural Networks), on DBrasiliensis dataset in order to compare them under both cyst detection and counting tasks.
\n The Dendrocephalus brasiliensis, a native species from South America, is a freshwater crustacean well explored in conservational and productive activities. Its main characteristics are its rusticity and resistance cysts production, in which the hatching requires a period of dehydration. Independent of the species utilization nature, it is essential to manipulate its cysts, such as the counting using microscopes. Manually counting is a difficult task, prone to errors, and that also very time-consuming. In this paper, we propose an automatized approach for the detection and counting of Dendrocephalus brasiliensis cysts from images captured by a digital microscope. For this purpose, we built the DBrasiliensis dataset, a repository with 246 images containing 5141 cysts of Dendrocephalus brasiliensis. Then, we trained two state-of-the-art object detection methods, YOLOv3 (You Only Look Once) and Faster R-CNN (Region-based Convolutional Neural Networks), on DBrasiliensis dataset in order to compare them under both cyst detection and counting tasks. Experiments showed evidence that YOLOv3 is superior to Faster R-CNN, achieving an accuracy rate of 83,74%, R2 of 0.88, RMSE (Root Mean Square Error) of 3.49, and MAE (Mean Absolute Error) of 2.24 on cyst detection and counting. Moreover, we showed that is possible to infer the number of cysts of a substrate, with known weight, by performing the automated counting of some of its samples. In conclusion, the proposed approach using YOLOv3 is adequate to detect and count Dendrocephalus brasiliensis cysts. The DBrasiliensis dataset can be accessed at: https://doi.org/10.6084/m9.figshare.13073240.

\n
\n
\n

SpikeSegNet-a deep learning approach utilizing encoder-decoder network with hourglass for spike segmentation and counting in wheat plant from visual imaging
\n Background High throughput non-destructive phenotyping is emerging as a significant approach for phenotyping germplasm and breeding populations for the identification of superior donors, elite lines, and QTLs. Detection and counting of spikes, the grain bearing organs of wheat, is critical for phenomics of a large set of germplasm and breeding lines in controlled and field conditions. It is also required for precision agriculture where the application of nitrogen, water, and other inputs at this critical stage is necessary. Further, counting of spikes is an important measure to determine yield. Digital image analysis and machine learning techniques play an essential role in non-destructive plant phenotyping analysis. Results In this study, an approach based on computer vision, particularly object detection, to recognize and count the number of spikes of the wheat plant from the digital images is proposed. For spike identification, a novel deep-learning network, SpikeSegNet, has been developed by combining two proposed feature networks: Local Patch extraction Network (LPNet) and Global Mask refinement Network (GMRNet). In LPNet, the contextual and spatial features are learned at the local patch level. The output of LPNet is a segmented mask image, which is further refined at the global level using GMRNet. Visual (RGB) images of 200 wheat plants were captured using LemnaTec imaging system installed at Nanaji Deshmukh Plant Phenomics Centre, ICAR-IARI, New Delhi. The precision, accuracy, and robustness (F 1 score) of the proposed approach for spike segmentation are found to be 99.93%, 99.91%, and 99.91%, respectively. For counting the number of spikes, “analyse particles”—function of imageJ was applied on the output image of the proposed SpikeSegNet model. For spike counting, the average precision, accuracy, and robustness are 99%, 95%, and 97%, respectively. SpikeSegNet approach is tested for robustness with illuminated image dataset, and no significant difference is observed in the segmentation performance. Conclusion In this study, a new approach called as SpikeSegNet has been proposed based on combined digital image analysis and deep learning techniques. A dedicated deep learning approach has been developed to identify and count spikes in the wheat plants. The performance of the approach demonstrates that SpikeSegNet is an effective and robust approach for spike detection and counting. As detection and counting of wheat spikes are closely related to the crop yield, and the proposed approach is also non-destructive, it is a significant step forward in the area of non-destructive and high-throughput phenotyping of wheat.

\n
\n
\n

A Simple Vehicle Counting System Using Deep Learning with YOLOv3 Model
\n Deep Learning is a popular Machine Learning algorithm that is widely used in many areas in current daily life. Its robust performance and ready-to-use frameworks and architectures enables many people to develop various Deep Learning-based software or systems to support human tasks and activities. Traffic monitoring is one area that utilizes Deep Learning for several purposes. By using cameras installed in some spots on the roads, many tasks such as vehicle counting, vehicle identification, traffic violation monitoring, vehicle speed monitoring, etc. can be realized. In this paper, we discuss a Deep Learning implementation to create a vehicle counting system without having to track the vehicles movements. To enhance the system performance and to reduce time in deploying Deep Learning architecture, hence pretrained model of YOLOv3 is used in this research due to its good performance and moderate computational time in object detection. This research aims to create a simple vehicle counting system to help human in classify and counting the vehicles that cross the street. The counting is based on four types of vehicle, i.e. car, motorcycle, bus, and truck, while previous research counts the car only. As the result, our proposed system capable to count the vehicles crossing the road based on video captured by camera with the highest accuracy of 97.72%.

\n
\n
\n

Accurate stacked-sheet counting method based on deep learning.
\n TL;DR: This paper proposes a new stacked-sheet counting method with a deep learning approach using the U-Net architecture. A specific dataset according to the characteristics of stack side images is collected. The stripe of the center line of each sheet is used for semantic segmentation, and the complete side images of the slices are segmented via training with small image patches and testing with original large images. With this model, each pixel is classified by multi-layer convolution and deconvolution to determine whether it is the target object to be detected. After the model is trained, the test set is used to test the model, and a center region
\n The accurate counting of laminated sheets, such as packing or printing sheets in industry, is extremely important because it greatly affects the economic cost. However, the different thicknesses, adhesion properties, and breakage points and the low contrast of sheets remain challenges to traditional counting methods based on image processing. This paper proposes a new stacked-sheet counting method with a deep learning approach using the U-Net architecture. A specific dataset according to the characteristics of stack side images is collected. The stripe of the center line of each sheet is used for semantic segmentation, and the complete side images of the slices are segmented via training with small image patches and testing with original large images. With this model, each pixel is classified by multi-layer convolution and deconvolution to determine whether it is the target object to be detected. After the model is trained, the test set is used to test the model, and a center region segmentation map based on the pixel points is obtained. By calculating the statistical median value of centerline points across different sections in these segmented images, the number of sheets can be obtained. Compared with traditional image algorithms in real product counting experiments, the proposed method can achieve better performance with higher accuracy and a lower error rate.

\n
\n
\n

Automatic Detection and Counting of Lymphocytes from Immunohistochemistry Cancer Images Using Deep Learning
\n Purpose Cancer is one of the most life-threatening and devastating diseases in the world. The generally recognized standard for cancer staging is the TNM staging system. In addition, a new parameter called the Immunoscore has been developed recently to assess the survival rate of patients. The Immunoscore is based on counts of CD3+ and CD8+ lymphocytes in the tumour core and margin. Counting the number of lymphocytes is a tedious process for pathologists. This paper examines the use of deep learning techniques for automatic detection and counting of lymphocytes from immunohistochemistry images of breast, colon, and prostate cancers. Methods We used an object detector called Faster R-CNN with four feature extractors: Resnet-50, VGG-16, Inception-V2, and Resnet-101 for automatic detection and counting of lymphocytes. A total of 11,136 lymphocytes were annotated after performing data augmentation on 1228 images. The test images are separated into three regions of interest (ROI): scattered lymphocytes, groups of lymphocytes, and artefacts. In each ROI, the performance of the object detector was checked by evaluation metrics. Results On comparing the F1-score for all three ROIs, we found that Resnet-101 provides better performance than the other feature extractors when using Faster R-CNN. The mean error in lymphocyte count for all ROIs appears to be insignificant. The detection time for a single image is less for VGG-16 than for other feature extractors. Conclusion This study presents a fine-tuned Faster R-CNN object detector for automatic detection and counting of lymphocytes in three different cancer tissues for scoring lymphocytes. Our results suggest that the Faster R-CNN method is efficient and yields good results. Thus, the proposed method can assist pathologists in providing a cancer prognosis.

\n
\n
", "rawMarkdown": "\n\n'\n\nThis is a list of papers that will make you smarter. \nThis is a list of papers that will make you more thoughtful. \nThis is a list of papers that will make you analyze the world in new ways. \nThis is a list of papers that will make you see the world in a different light.\n\nThis is a list of papers that might make you get a better score.\nBecause this is what we are here for, right ;)\n\n\nEnjoy! <3\n\n_____\n**[An Advanced Deep Learning Approach for Multi-Object Counting in Urban Vehicular Environments](https://doi.org/10.3390/fi13120306)**\n\nObject counting is an active research area that gained more attention in the past few years. In smart cities, vehicle counting plays a crucial role in urban planning and management of the Intelligent Transportation Systems (ITS). Several approaches have been proposed in the literature to address this problem. However, the resulting detection accuracy is still not adequate. This paper proposes an efficient approach that uses deep learning concepts and correlation filters for multi-object counting and tracking. The performance of the proposed system is evaluated using a dataset consisting of 16 videos with different features to examine the impact of object density, image quality, angle of view, and speed of motion towards system accuracy. Performance evaluation exhibits promising results in normal traffic scenarios and adverse weather conditions. Moreover, the proposed approach outperforms the performance of two recent approaches from the literature.\n_____\n\n_____\n**[Object Counting using Deep Learning](https://doi.org/10.17762/TURCOMAT.V12I10.4432)**\n\nIn this paper, we consider the Problem of Object counting in deep learning. It is frequently carried out in different place of industries, school and colleges, traffic places among others. Object counting is major for quantitative analyses that rely on evaluation on certain objects. In this work, we propose a deep learning to find this challenge. Unfortunately, Object counting is most commonly a manual task and can be time intensive.  As a result, we manage both to increase the accuracy count and decrease the processing time. A Deep Learning based system can be used for real time applications.\n_____\n\n_____\n**[Towards Perspective-Free Object Counting with Deep Learning](https://doi.org/10.1007/978-3-319-46478-7_38)**\n\n\n> **TL;DR:** Our models are able to precisely estimate the number of vehicles in a traffic congestion, or to count the humans in a very crowded scene. Our first contribution is the proposal of a novel convolutional neural network solution, named Counting CNN (CCNN). Essentially, the CCNN is formulated as a regression model where the network learns how to map the appearance of the image patches to their corresponding object density maps. Our second contribution consists in a scale-aware counting model, the Hydra CNN, able to estimate object densities in different very crowded scenarios where no geometric information of the scene can be provided. Hydra CNN learns a multisc\n\nIn this paper we address the problem of counting objects instances in images. Our models are able to precisely estimate the number of vehicles in a traffic congestion, or to count the humans in a very crowded scene. Our first contribution is the proposal of a novel convolutional neural network solution, named Counting CNN (CCNN). Essentially, the CCNN is formulated as a regression model where the network learns how to map the appearance of the image patches to their corresponding object density maps. Our second contribution consists in a scale-aware counting model, the Hydra CNN, able to estimate object densities in different very crowded scenarios where no geometric information of the scene can be provided. Hydra CNN learns a multiscale non-linear regression model which uses a pyramid of image patches extracted at multiple scales to perform the final density prediction. We report an extensive experimental evaluation, using up to three different object counting benchmarks, where we show how our solutions achieve a state-of-the-art performance.\n_____\n\n_____\n**[Everything counts: a Taxonomy of Deep Learning Approaches for Object Counting](https://scholar.google.com/scholar?q=Everything%20counts%3A%20a%20Taxonomy%20of%20Deep%20Learning%20Approaches%20for%20Object%20Counting)**\n\n\n> **TL;DR:** We provide a taxonomy that reflects both: the method characteristic as well as the counting problem characteristics in order to derive a systematization of the scientific field. This taxonomy is applied on four case studies.\n\nMany tasks, like process fault detection, disease diagnostic or maintaining security in public places are connected to the process of counting objects from observing image data like camera footage or microscopic images. However, the pivotal process of counting objects based on image data is subject to a lot of error sources when done manually. Therefore, a lot of approaches exist that aim to aid the automation of the counting process. Since most of those methods are of black box nature and have to be applied to real world scenarios, we provide a taxonomy that reflects both: the method characteristic as well as the counting problem characteristics in order to derive a systematization of the scientific field and to provide a tool that acts as a set of guidelines for choosing appropriate counting methods for different situations. We first conduct a literature review, where counting problem characteristics and solutions are extracted and later transformed into a taxonomy. We finally showcase the taxonomy using four case studies that are based on publicly available datasets for the sake of scientific comprehension.\n_____\n\n_____\n**[Crowdsourcing annotation system of object counting dataset for deep learning algorithm](https://doi.org/10.1088/1755-1315/195/1/012063)**\n\nDeep Learning is currently the state-of-the-art technique for various Computer Vision tasks, including object counting. Despite of its high performance, Deep Learning requires a gigantic amount of training data to show its best result. Getting this massive data in reasonable time requires a proper strategy such as crowdsourcing. However, in case of object counting, we found no crowdsourcing system able to effectively collect necessary data. To tackle this problem, we develop a crowdsourcing system to annotate image for object counting dataset. This system is also equipped with validation system to ensure the quality of collected dataset.\n_____\n\n_____\n**[Automatic Pest Counting from Pheromone Trap Images Using Deep Learning Object Detectors for Matsucoccus thunbergianae Monitoring](https://doi.org/10.3390/insects12040342)**\n\nSimple Summary The black pine bast scale, Matsucoccus thunbergianae, is a forest pest that causes widespread damage to black pine; therefore, monitoring this pest is necessary to minimize environmental and economic losses in forests. However, monitoring insects in pheromone traps performed by humans is labor intensive and time consuming. To develop an automated monitoring system, we aimed to develop algorithms that detect and count M. thunbergianae from images of pheromone traps using deep-learning-based object detection algorithms. Object detection models based on deep learning neural networks under various conditions were trained, and the performances of detection and counting were compared and evaluated. In addition, the models were trained to detect small objects well by cropping images into multiple windows. As a result, the algorithms based on deep learning neural networks successfully detected and counted M. thunbergianae. These results showed that accurate and constant pest monitoring is possible using the artificial-intelligence-based methods we proposed. Abstract The black pine bast scale, M. thunbergianae, is a major insect pest of black pine and causes serious environmental and economic losses in forests. Therefore, it is essential to monitor the occurrence and population of M. thunbergianae, and a monitoring method using a pheromone trap is commonly employed. Because the counting of insects performed by humans in these pheromone traps is labor intensive and time consuming, this study proposes automated deep learning counting algorithms using pheromone trap images. The pheromone traps collected in the field were photographed in the laboratory, and the images were used for training, validation, and testing of the detection models. In addition, the image cropping method was applied for the successful detection of small objects in the image, considering the small size of M. thunbergianae in trap images. The detection and counting performance were evaluated and compared for a total of 16 models under eight model conditions and two cropping conditions, and a counting accuracy of 95% or more was shown in most models. This result shows that the artificial intelligence-based pest counting method proposed in this study is suitable for constant and accurate monitoring of insect pests.\n_____\n\n_____\n**[Counting Dense Leaves under Natural Environments via an Improved Deep-Learning-Based Object Detection Algorithm](https://doi.org/10.3390/agriculture11101003)**\n\nThe leaf is the organ that is crucial for photosynthesis and the production of nutrients in plants; as such, the number of leaves is one of the key indicators with which to describe the development and growth of a canopy. The irregular shape and distribution of the blades, as well as the effect of natural light, make the segmentation and detection process of the blades difficult. The inaccurate acquisition of plant phenotypic parameters may affect the subsequent judgment of crop growth status and crop yield. To address the challenge in counting dense and overlapped plant leaves under natural environments, we proposed an improved deep-learning-based object detection algorithm by merging a space-to-depth module, a Convolutional Block Attention Module (CBAM) and Atrous Spatial Pyramid Pooling (ASPP) into the network, and applying the smoothL1 function to improve the loss function of object prediction. We evaluated our method on images of five different plant species collected under indoor and outdoor environments. The experimental results demonstrated that our algorithm which counts dense leaves improved average detection accuracy of 85% to 96%. Our algorithm also showed better performance in both detection accuracy and time consumption compared to other state-of-the-art object detection algorithms.\n_____\n\n_____\n**[Plant Counting of Cotton from UAS Imagery Using Deep Learning-Based Object Detection Framework](https://doi.org/10.3390/rs12182981)**\n\nAssessing plant population of cotton is important to make replanting decisions in low plant density areas, prone to yielding penalties. Since the measurement of plant population in the field is labor intensive and subject to error, in this study, a new approach of image-based plant counting is proposed, using unmanned aircraft systems (UAS; DJI Mavic 2 Pro, Shenzhen, China) data. The previously developed image-based techniques required a priori information of geometry or statistical characteristics of plant canopy features, while also limiting the versatility of the methods in variable field conditions. In this regard, a deep learning-based plant counting algorithm was proposed to reduce the number of input variables, and to remove requirements for acquiring geometric or statistical information. The object detection model named You Only Look Once version 3 (YOLOv3) and photogrammetry were utilized to separate, locate, and count cotton plants in the seedling stage. The proposed algorithm was tested with four different UAS datasets, containing variability in plant size, overall illumination, and background brightness. Root mean square error (RMSE) and R2 values of the optimal plant count results ranged from 0.50 to 0.60 plants per linear meter of row (number of plants within 1 m distance along the planting row direction) and 0.96 to 0.97, respectively. The object detection algorithm, trained with variable plant size, ground wetness, and lighting conditions generally resulted in a lower detection error, unless an observable difference of developmental stages of cotton existed. The proposed plant counting algorithm performed well with 0–14 plants per linear meter of row, when cotton plants are generally separable in the seedling stage. This study is expected to provide an automated methodology for in situ evaluation of plant emergence using UAS data.\n_____\n\n_____\n**[Vehicle Counting System using Deep Learning and Multi-Object Tracking Methods](https://doi.org/10.1177/0361198120912742)**\n\nUsing deep learning technology and multi-object tracking method to count vehicles accurately in different traffic conditions is a hot research topic in the field of intelligent transportation. In this paper, first, a vehicle dataset from the perspective of highway surveillance cameras is constructed, and the vehicle detection model is obtained by training using the You Only Look Once (YOLO) version 3 network. Second, an improved multi-scale and multi-feature tracking algorithm based on a kernel correlation filter (KCF) algorithm is proposed to avoid the KCF extracting single features and single-scale defects. Combining the intersection over union (IoU) similarity measure and the row-column optimal association criterion proposed in this paper, matching strategy is used to process the vehicles that are not detected and wrongly detected, thereby obtaining complete vehicle trajectories. Finally, according to the trajectory of the vehicle, the traveling direction of the vehicle is automatically determined, and the setting position of the detecting line is automatically updated to obtain the vehicle count result accurately. Experiments were conducted in a variety of traffic scenes and compared with published data. The experimental results show that the proposed method achieves high accuracy of vehicle detection while maintaining accuracy and precision in tracking multiple objects, and obtains accurate vehicle counting results which can meet real-time processing requirements. The algorithm presented in this paper has practical application for vehicle counting in complex highway scenes.\n_____\n\n_____\n**[Occlusion Robust Wheat Ear Counting Algorithm Based on Deep Learning](https://doi.org/10.3389/fpls.2021.645899)**\n\nCounting the number of wheat ears in images under natural light is an important way to evaluate the crop yield, thus, it is of great significance to modern intelligent agriculture. However, the distribution of wheat ears is dense, so the occlusion and overlap problem appears in almost every wheat image. It is difficult for traditional image processing methods to solve occlusion problem due to the deficiency of high-level semantic features, while existing deep learning based counting methods did not solve the occlusion efficiently. This article proposes an improved EfficientDet-D0 object detection model for wheat ear counting, and focuses on solving occlusion. First, the transfer learning method is employed in the pre-training of the model backbone network to extract the high-level semantic features of wheat ears. Secondly, an image augmentation method Random-Cutout is proposed, in which some rectangles are selected and erased according to the number and size of the wheat ears in the images to simulate occlusion in real wheat images. Finally, convolutional block attention module (CBAM) is adopted into the EfficientDet-D0 model after the backbone, which makes the model refine the features, pay more attention to the wheat ears and suppress other useless background information. Extensive experiments are done by feeding the features to detection layer, showing that the counting accuracy of the improved EfficientDet-D0 model reaches 94%, which is about 2% higher than the original model, and false detection rate is 5.8%, which is the lowest among comparative methods.\n_____\n\n_____\n**[Applications of object detection in modular construction based on a comparative evaluation of deep learning algorithms](https://doi.org/10.1108/CI-02-2020-0017)**\n\n\nPurpose\nThe practice of artificial intelligence (AI) is increasingly being promoted by technology developers. However, its adoption rate is still reported as low in the construction industry due to a lack of expertise and the limited reliable applications for AI technology. Hence, this paper aims to present the detailed outcome of experimentations evaluating the applicability and the performance of AI object detection algorithms for construction modular object detection.\n\n\nDesign/methodology/approach\nThis paper provides a thorough evaluation of two deep learning algorithms for object detection, including the faster region-based convolutional neural network (faster RCNN) and single shot multi-box detector (SSD). Two types of metrics are also presented; first, the average recall and mean average precision by image pixels; second, the recall and precision by counting. To conduct the experiments using the selected algorithms, four infrastructure and building construction sites are chosen to collect the required data, including a total of 990 images of three different but common modular objects, including modular panels, safety barricades and site fences.\n\n\nFindings\nThe results of the comprehensive evaluation of the algorithms show that the performance of faster RCNN and SSD depends on the context that detection occurs. Indeed, surrounding objects and the backgrounds of the objects affect the level of accuracy obtained from the AI analysis and may particularly effect precision and recall. The analysis of loss lines shows that the loss lines for selected objects depend on both their geometry and the image background. The results on selected objects show that faster RCNN offers higher accuracy than SSD for detection of selected objects.\n\n\nResearch limitations/implications\nThe results show that modular object detection is crucial in construction for the achievement of the required information for project quality and safety objectives. The detection process can significantly improve monitoring object installation progress in an accurate and machine-based manner avoiding human errors. The results of this paper are limited to three construction sites, but future investigations can cover more tasks or objects from different construction sites in a fully automated manner.\n\n\nOriginality/value\nThis paper’s originality lies in offering new AI applications in modular construction, using a large first-hand data set collected from three construction sites. Furthermore, the paper presents the scientific evaluation results of implementing recent object detection algorithms across a set of extended metrics using the original training and validation data sets to improve the generalisability of the experimentation. This paper also provides the practitioners and scholars with a workflow on AI applications in the modular context and the first-hand referencing data.\n\n_____\n\n_____\n**[Leaf Counting in Rice (Oryza Sativa L.) Using Object Detection: A Deep Learning Approach](https://doi.org/10.1109/IGARSS39084.2020.9324153)**\n\nLeaf count is one of the crucial tasks in plant phenotyping, and leaves are the basic unit of plant architecture involved in photosynthesis, growth, and yield of a plant. Therefore, the total number of leaves per plant is considered as one of the essential physio-morphological plant traits for phenotyping. The current work proposes to estimate the total number of leaves of a rice plant by detecting their leaves tips. A rice plant has a single tip for a single leaf. Hence, this proposed framework counts the total number of leaves by counting the number of leaves tips equal to the number of leaves. You Only Look Once (YOLO) algorithm is used for the detection of the leaves tips as an object. This hypothesis builds a basis for counting the total number of leaves in a plant like rice, and similar field crops such as wheat (Triticum aestivum L), maize (Zea mays L.), sorghum (Sorghum bicolor), barley (Hordeum vulgare L.). The model detected leaves of a rice plant (RGB images) by detecting corresponding leaves tips with YOLO having average accuracy up to 82% and IOU around 0.53-0.60 and estimates the number of leaves in a plant by counting predicted bounding boxes around tips. The model also performed well with the wheat crop.\n_____\n\n_____\n**[Central Object Segmentation by Deep Learning to Continuously Monitor Fruit Growth through RGB Images](https://doi.org/10.3390/s21216999)**\n\nMonitoring fruit growth is useful when estimating final yields in advance and predicting optimum harvest times. However, observing fruit all day at the farm via RGB images is not an easy task because the light conditions are constantly changing. In this paper, we present CROP (Central Roundish Object Painter). The method involves image segmentation by deep learning, and the architecture of the neural network is a deeper version of U-Net. CROP identifies different types of central roundish fruit in an RGB image in varied light conditions, and creates a corresponding mask. Counting the mask pixels gives the relative two-dimensional size of the fruit, and in this way, time-series images may provide a non-contact means of automatically monitoring fruit growth. Although our measurement unit is different from the traditional one (length), we believe that shape identification potentially provides more information. Interestingly, CROP can have a more general use, working even for some other roundish objects. For this reason, we hope that CROP and our methodology yield big data to promote scientific advancements in horticultural science and other fields.\n_____\n\n_____\n**[A NOVEL DEEP LEARNING BASED METHOD FOR DETECTION AND COUNTING OF VEHICLES IN URBAN TRAFFIC SURVEILLANCE SYSTEMS](https://doi.org/10.5194/isprs-archives-xliii-b2-2021-793-2021)**\n\nIn intelligent transportation systems (ITS), it is essential to obtain reliable statistics of the vehicular flow in order to create urban traffic management strategies. These systems have benefited from the increase in computational resources and the improvement of image processing methods, especially in object detection based on deep learning. This paper proposes a method for vehicle counting composed of three stages: object detection, tracking and trajectory processing. In order to select the detection model with the best trade-off between accuracy and speed, the following one-stage detection models were compared: SSD512, CenterNet, Efficiedet-D0 and YOLO family models (v2, v3 and v4). Experimental results conducted on the benchmark dataset show that the best rates among the detection models were obtained using YOLOv4 with mAP =87% and a processing speed of 18 FPS. On the other hand, the accuracy obtained in the proposed counting method was 94% with a real-time processing rate lower than 1.9.\n_____\n\n_____\n**[YOLO-Based Deep Learning Framework for Olive Fruit Fly Detection and Counting](https://doi.org/10.1109/ACCESS.2021.3088075)**\n\nThe olive fruit fly can damage up to 100% of the harvested fruit and can cause up to 80% reduction of the value of the resulting olive oil. Therefore, it is important to early detect its presence in the olive orchard to take the appropriate chemical or biological countermeasures as early as possible. Traps filled with attractant pheromones are typically deployed across the orchard to attract and capture the flies. Traditionally, the captured flies were manually counted which is error prone. Recently, the traps are employed with cameras and communication devices to send pictures of the captured flies to experts for analysis which is also error prone and inefficient. Consequently, machine and deep learning have been exploited to develop fully automated and accurate detection that does not include human in the loop. Such a learning problem is challenging due to the small size of the detected object, the differences in the light conditions at which pictures were taken, and the lack of enough data to train the learning model. In this paper, we present a deep learning framework for detecting and counting the number of olive fruit flies that exploits data augmentation to increase the dataset size, includes negative samples in the training to improve the detection accuracy, and normalizes the images to the color of the trap background, i.e., yellow, to unify the illumination conditions. The results of the proposed framework show a precision of 0.84, a recall of 0.97, an F1-score of 0.9 and mean Average Precision (mAP) of 96.68% which significantly outperforms existing pest detection systems.\n_____\n\n_____\n**[A Deep Learning Bidirectional Temporal Tracking Algorithm for Automated Blood Cell Counting from Non-invasive Capillaroscopy Videos](https://doi.org/10.1007/978-3-030-87237-3_40)**\n\n\n> **TL;DR:** We propose a deep learning multi-cell tracking model, CycleTrack, which achieves accurate blood cell counting from capillaroscopic videos. Our approach combines two simple online tracking models, SORT and CenterTrack, and is tailored to features of capillary blood cell flow. Blood cells are tracked by displacement vectors in two opposing temporal directions (forward- and backward-tracking) between consecutive frames. The proposed model outperforms other baseline trackers, achieving 65.57% Multiple Object Tracking Accuracy and 73.95% ID F1 score on test videos. Compared to manual blood cell counting, CycleTrack achieves 96.58 $\\pm$\n\nOblique back-illumination capillaroscopy has recently been introduced as a method for high-quality, non-invasive blood cell imaging in human capillaries. To make this technique practical for clinical blood cell counting, solutions for automatic processing of acquired videos are needed. Here, we take the first step towards this goal, by introducing a deep learning multi-cell tracking model, named CycleTrack, which achieves accurate blood cell counting from capillaroscopic videos. CycleTrack combines two simple online tracking models, SORT and CenterTrack, and is tailored to features of capillary blood cell flow. Blood cells are tracked by displacement vectors in two opposing temporal directions (forward- and backward-tracking) between consecutive frames. This approach yields accurate tracking despite rapidly moving and deforming blood cells. The proposed model outperforms other baseline trackers, achieving 65.57% Multiple Object Tracking Accuracy and 73.95% ID F1 score on test videos. Compared to manual blood cell counting, CycleTrack achieves 96.58 $\\pm$ 2.43% cell counting accuracy among 8 test videos with 1000 frames each compared to 93.45% and 77.02% accuracy for independent CenterTrack and SORT almost without additional time expense. It takes 800s to track and count approximately 8000 blood cells from 9,600 frames captured in a typical one-minute video. Moreover, the blood cell velocity measured by CycleTrack demonstrates a consistent, pulsatile pattern within the physiological range of heart rate. Lastly, we discuss future improvements for the CycleTrack framework, which would enable clinical translation of the oblique back-illumination microscope towards a real-time and non-invasive point-of-care blood cell counting and analyzing technology.\n_____\n\n_____\n**[A NOVEL DEEP LEARNING BASED METHOD FOR DETECTION AND COUNTING OF VEHICLES IN URBAN TRAFFIC SURVEILLANCE SYSTEMS](https://doi.org/10.5194/isprs-archives-xliii-b2-2021-793-2021)**\n\n\n> **TL;DR:** This paper proposes a method for vehicle counting composed of three stages: object detection, tracking and trajectory processing. In order to select the detection model with the best trade-off between accuracy and speed, the following one-stage detection models were compared: SSD512, CenterNet, Efficiedet-D0 and YOLO family models (v2, v3 and v4). Experimental results conducted on the benchmark dataset show that the best rates among the detection models were obtained using YOLOv4 with mAP =87% and a processing speed of 18 FPS. On the other hand, the accuracy obtained in the\n\nIn intelligent transportation systems (ITS), it is essential to obtain reliable statistics of the vehicular flow in order to create urban traffic management strategies. These systems have benefited from the increase in computational resources and the improvement of image processing methods, especially in object detection based on deep learning. This paper proposes a method for vehicle counting composed of three stages: object detection, tracking and trajectory processing. In order to select the detection model with the best trade-off between accuracy and speed, the following one-stage detection models were compared: SSD512, CenterNet, Efficiedet-D0 and YOLO family models (v2, v3 and v4). Experimental results conducted on the benchmark dataset show that the best rates among the detection models were obtained using YOLOv4 with mAP =87% and a processing speed of 18 FPS. On the other hand, the accuracy obtained in the proposed counting method was 94% with a real-time processing rate lower than 1.9.\n_____\n\n_____\n**[Multiple Object Tracking using Deep Learning with YOLO V5](https://scholar.google.com/scholar?q=Multiple%20Object%20Tracking%20using%20Deep%20Learning%20with%20YOLO%20V5)**\n\nThe MOT (Multiple Object Tracking) is an important tool in the modern world. It has various uses like object detection, counting objects, security tools ,etc. The Object tracking is a prominent technology in image processing which has a large future scope. The MOT has made significant growth in a few years due to deep learning, computer vision, machine learning, etc. This paper aims to provide a software solution that keeps track of the objects so that it can handle object list and count. By using YOLO “You Only Look Once” Technology with the help of Pytorch, the system aims in object detection, tracking and counting. Also unlike the general yolo object detection tool which detects all objects at the same time ,this MOT system also detects only objects which are needed to be detected by the user and thus helps in improving the performance of the system.\n_____\n\n_____\n**[Tomato Fruit Detection and Counting in Greenhouses Using Deep Learning](https://doi.org/10.3389/fpls.2020.571299)**\n\nAccurately detecting and counting fruits during plant growth using imaging and computer vision is of importance not only from the point of view of reducing labor intensive manual measurements of phenotypic information, but also because it is a critical step toward automating processes such as harvesting. Deep learning based methods have emerged as the state-of-the-art techniques in many problems in image segmentation and classification, and have a lot of promise in challenging domains such as agriculture, where they can deal with the large variability in data better than classical computer vision methods. This paper reports results on the detection of tomatoes in images taken in a greenhouse, using the MaskRCNN algorithm, which detects objects and also the pixels corresponding to each object. Our experimental results on the detection of tomatoes from images taken in greenhouses using a RealSense camera are comparable to or better than the metrics reported by earlier work, even though those were obtained in laboratory conditions or using higher resolution images. Our results also show that MaskRCNN can implicitly learn object depth, which is necessary for background elimination.\n_____\n\n_____\n**[Towards a low-cost embedded vehicle counting system based on deep-learning for traffic management applications](https://doi.org/10.1109/CHILECON54041.2021.9702914)**\n\n\n> **TL;DR:** The Kalman filter is more efficient than the centroid tracking algorithm, and it produces fewer errors.\n\nThis paper explores the feasibility of using a low-cost embedded system for real-time vehicle detection and counting through the use of deep neural networks. It compares the performance of two different object tracking methods, the Kalman filter with the Hungarian algorithm and the centroid tracking algorithm. The experimentation proved that the efficiency of the implemented algorithms was above the 92% and 98% for the centroid tracking algorithm and Kalman filter with the Hungarian algorithm, respectively. Also, the Kalman filter produced fewer errors overcoming the centroid tracking algorithm.\n_____\n\n_____\n**[Recognizing and counting Dendrocephalus brasiliensis (Crustacea: Anostraca) cysts using deep learning](https://doi.org/10.1371/journal.pone.0248574)**\n\n\n> **TL;DR:** We propose an automatized approach for the detection and counting of Dendrocephalus brasiliensis cysts from images captured by a digital microscope. For this purpose, we built the DBrasiliensis dataset, a repository with 246 images containing 5141 cysts of Dendrocephalus brasiliensis. Then, we trained two state-of-the-art object detection methods, YOLOv3 (You Only Look Once) and Faster R-CNN (Region-based Convolutional Neural Networks), on DBrasiliensis dataset in order to compare them under both cyst detection and counting tasks.\n\nThe Dendrocephalus brasiliensis, a native species from South America, is a freshwater crustacean well explored in conservational and productive activities. Its main characteristics are its rusticity and resistance cysts production, in which the hatching requires a period of dehydration. Independent of the species utilization nature, it is essential to manipulate its cysts, such as the counting using microscopes. Manually counting is a difficult task, prone to errors, and that also very time-consuming. In this paper, we propose an automatized approach for the detection and counting of Dendrocephalus brasiliensis cysts from images captured by a digital microscope. For this purpose, we built the DBrasiliensis dataset, a repository with 246 images containing 5141 cysts of Dendrocephalus brasiliensis. Then, we trained two state-of-the-art object detection methods, YOLOv3 (You Only Look Once) and Faster R-CNN (Region-based Convolutional Neural Networks), on DBrasiliensis dataset in order to compare them under both cyst detection and counting tasks. Experiments showed evidence that YOLOv3 is superior to Faster R-CNN, achieving an accuracy rate of 83,74%, R2 of 0.88, RMSE (Root Mean Square Error) of 3.49, and MAE (Mean Absolute Error) of 2.24 on cyst detection and counting. Moreover, we showed that is possible to infer the number of cysts of a substrate, with known weight, by performing the automated counting of some of its samples. In conclusion, the proposed approach using YOLOv3 is adequate to detect and count Dendrocephalus brasiliensis cysts. The DBrasiliensis dataset can be accessed at: https://doi.org/10.6084/m9.figshare.13073240.\n_____\n\n_____\n**[SpikeSegNet-a deep learning approach utilizing encoder-decoder network with hourglass for spike segmentation and counting in wheat plant from visual imaging](https://doi.org/10.1186/s13007-020-00582-9)**\n\nBackground High throughput non-destructive phenotyping is emerging as a significant approach for phenotyping germplasm and breeding populations for the identification of superior donors, elite lines, and QTLs. Detection and counting of spikes, the grain bearing organs of wheat, is critical for phenomics of a large set of germplasm and breeding lines in controlled and field conditions. It is also required for precision agriculture where the application of nitrogen, water, and other inputs at this critical stage is necessary. Further, counting of spikes is an important measure to determine yield. Digital image analysis and machine learning techniques play an essential role in non-destructive plant phenotyping analysis. Results In this study, an approach based on computer vision, particularly object detection, to recognize and count the number of spikes of the wheat plant from the digital images is proposed. For spike identification, a novel deep-learning network, SpikeSegNet, has been developed by combining two proposed feature networks: Local Patch extraction Network (LPNet) and Global Mask refinement Network (GMRNet). In LPNet, the contextual and spatial features are learned at the local patch level. The output of LPNet is a segmented mask image, which is further refined at the global level using GMRNet. Visual (RGB) images of 200 wheat plants were captured using LemnaTec imaging system installed at Nanaji Deshmukh Plant Phenomics Centre, ICAR-IARI, New Delhi. The precision, accuracy, and robustness (F 1 score) of the proposed approach for spike segmentation are found to be 99.93%, 99.91%, and 99.91%, respectively. For counting the number of spikes, “analyse particles”—function of imageJ was applied on the output image of the proposed SpikeSegNet model. For spike counting, the average precision, accuracy, and robustness are 99%, 95%, and 97%, respectively. SpikeSegNet approach is tested for robustness with illuminated image dataset, and no significant difference is observed in the segmentation performance. Conclusion In this study, a new approach called as SpikeSegNet has been proposed based on combined digital image analysis and deep learning techniques. A dedicated deep learning approach has been developed to identify and count spikes in the wheat plants. The performance of the approach demonstrates that SpikeSegNet is an effective and robust approach for spike detection and counting. As detection and counting of wheat spikes are closely related to the crop yield, and the proposed approach is also non-destructive, it is a significant step forward in the area of non-destructive and high-throughput phenotyping of wheat.\n_____\n\n_____\n**[A Simple Vehicle Counting System Using Deep Learning with YOLOv3 Model](https://doi.org/10.29207/resti.v4i3.1871)**\n\nDeep Learning is a popular Machine Learning algorithm that is widely used in many areas in current daily life. Its robust performance and ready-to-use frameworks and architectures enables many people to develop various Deep Learning-based software or systems to support human tasks and activities. Traffic monitoring is one area that utilizes Deep Learning for several purposes. By using cameras installed in some spots on the roads, many tasks such as vehicle counting, vehicle identification, traffic violation monitoring, vehicle speed monitoring, etc. can be realized. In this paper, we discuss a Deep Learning implementation to create a vehicle counting system without having to track the vehicles movements. To enhance the system performance and to reduce time in deploying Deep Learning architecture, hence pretrained model of YOLOv3 is used in this research due to its good performance and moderate computational time in object detection. This research aims to create a simple vehicle counting system to help human in classify and counting the vehicles that cross the street. The counting is based on four types of vehicle, i.e. car, motorcycle, bus, and truck, while previous research counts the car only. As the result, our proposed system capable to count the vehicles crossing the road based on video captured by camera with the highest accuracy of 97.72%.\n_____\n\n_____\n**[Accurate stacked-sheet counting method based on deep learning.](https://doi.org/10.1364/josaa.387390)**\n\n\n> **TL;DR:** This paper proposes a new stacked-sheet counting method with a deep learning approach using the U-Net architecture. A specific dataset according to the characteristics of stack side images is collected. The stripe of the center line of each sheet is used for semantic segmentation, and the complete side images of the slices are segmented via training with small image patches and testing with original large images. With this model, each pixel is classified by multi-layer convolution and deconvolution to determine whether it is the target object to be detected. After the model is trained, the test set is used to test the model, and a center region\n\nThe accurate counting of laminated sheets, such as packing or printing sheets in industry, is extremely important because it greatly affects the economic cost. However, the different thicknesses, adhesion properties, and breakage points and the low contrast of sheets remain challenges to traditional counting methods based on image processing. This paper proposes a new stacked-sheet counting method with a deep learning approach using the U-Net architecture. A specific dataset according to the characteristics of stack side images is collected. The stripe of the center line of each sheet is used for semantic segmentation, and the complete side images of the slices are segmented via training with small image patches and testing with original large images. With this model, each pixel is classified by multi-layer convolution and deconvolution to determine whether it is the target object to be detected. After the model is trained, the test set is used to test the model, and a center region segmentation map based on the pixel points is obtained. By calculating the statistical median value of centerline points across different sections in these segmented images, the number of sheets can be obtained. Compared with traditional image algorithms in real product counting experiments, the proposed method can achieve better performance with higher accuracy and a lower error rate.\n_____\n\n_____\n**[Automatic Detection and Counting of Lymphocytes from Immunohistochemistry Cancer Images Using Deep Learning](https://doi.org/10.1007/s40846-020-00545-4)**\n\nPurpose Cancer is one of the most life-threatening and devastating diseases in the world. The generally recognized standard for cancer staging is the TNM staging system. In addition, a new parameter called the Immunoscore has been developed recently to assess the survival rate of patients. The Immunoscore is based on counts of CD3+ and CD8+ lymphocytes in the tumour core and margin. Counting the number of lymphocytes is a tedious process for pathologists. This paper examines the use of deep learning techniques for automatic detection and counting of lymphocytes from immunohistochemistry images of breast, colon, and prostate cancers. Methods We used an object detector called Faster R-CNN with four feature extractors: Resnet-50, VGG-16, Inception-V2, and Resnet-101 for automatic detection and counting of lymphocytes. A total of 11,136 lymphocytes were annotated after performing data augmentation on 1228 images. The test images are separated into three regions of interest (ROI): scattered lymphocytes, groups of lymphocytes, and artefacts. In each ROI, the performance of the object detector was checked by evaluation metrics. Results On comparing the F1-score for all three ROIs, we found that Resnet-101 provides better performance than the other feature extractors when using Faster R-CNN. The mean error in lymphocyte count for all ROIs appears to be insignificant. The detection time for a single image is less for VGG-16 than for other feature extractors. Conclusion This study presents a fine-tuned Faster R-CNN object detector for automatic detection and counting of lymphocytes in three different cancer tissues for scoring lymphocytes. Our results suggest that the Faster R-CNN method is efficient and yields good results. Thus, the proposed method can assist pathologists in providing a cancer prognosis.\n_____\n", "votes": 2 } ], "comments": [], "raw_markdown_by_id": { "1781810": "\n\n'\n\nThis is a list of papers that will make you smarter. \nThis is a list of papers that will make you more thoughtful. \nThis is a list of papers that will make you analyze the world in new ways. \nThis is a list of papers that will make you see the world in a different light.\n\nThis is a list of papers that might make you get a better score.\nBecause this is what we are here for, right ;)\n\n\nEnjoy! <3\n\n_____\n**[An Advanced Deep Learning Approach for Multi-Object Counting in Urban Vehicular Environments](https://doi.org/10.3390/fi13120306)**\n\nObject counting is an active research area that gained more attention in the past few years. In smart cities, vehicle counting plays a crucial role in urban planning and management of the Intelligent Transportation Systems (ITS). Several approaches have been proposed in the literature to address this problem. However, the resulting detection accuracy is still not adequate. This paper proposes an efficient approach that uses deep learning concepts and correlation filters for multi-object counting and tracking. The performance of the proposed system is evaluated using a dataset consisting of 16 videos with different features to examine the impact of object density, image quality, angle of view, and speed of motion towards system accuracy. Performance evaluation exhibits promising results in normal traffic scenarios and adverse weather conditions. Moreover, the proposed approach outperforms the performance of two recent approaches from the literature.\n_____\n\n_____\n**[Object Counting using Deep Learning](https://doi.org/10.17762/TURCOMAT.V12I10.4432)**\n\nIn this paper, we consider the Problem of Object counting in deep learning. It is frequently carried out in different place of industries, school and colleges, traffic places among others. Object counting is major for quantitative analyses that rely on evaluation on certain objects. In this work, we propose a deep learning to find this challenge. Unfortunately, Object counting is most commonly a manual task and can be time intensive.  As a result, we manage both to increase the accuracy count and decrease the processing time. A Deep Learning based system can be used for real time applications.\n_____\n\n_____\n**[Towards Perspective-Free Object Counting with Deep Learning](https://doi.org/10.1007/978-3-319-46478-7_38)**\n\n\n> **TL;DR:** Our models are able to precisely estimate the number of vehicles in a traffic congestion, or to count the humans in a very crowded scene. Our first contribution is the proposal of a novel convolutional neural network solution, named Counting CNN (CCNN). Essentially, the CCNN is formulated as a regression model where the network learns how to map the appearance of the image patches to their corresponding object density maps. Our second contribution consists in a scale-aware counting model, the Hydra CNN, able to estimate object densities in different very crowded scenarios where no geometric information of the scene can be provided. Hydra CNN learns a multisc\n\nIn this paper we address the problem of counting objects instances in images. Our models are able to precisely estimate the number of vehicles in a traffic congestion, or to count the humans in a very crowded scene. Our first contribution is the proposal of a novel convolutional neural network solution, named Counting CNN (CCNN). Essentially, the CCNN is formulated as a regression model where the network learns how to map the appearance of the image patches to their corresponding object density maps. Our second contribution consists in a scale-aware counting model, the Hydra CNN, able to estimate object densities in different very crowded scenarios where no geometric information of the scene can be provided. Hydra CNN learns a multiscale non-linear regression model which uses a pyramid of image patches extracted at multiple scales to perform the final density prediction. We report an extensive experimental evaluation, using up to three different object counting benchmarks, where we show how our solutions achieve a state-of-the-art performance.\n_____\n\n_____\n**[Everything counts: a Taxonomy of Deep Learning Approaches for Object Counting](https://scholar.google.com/scholar?q=Everything%20counts%3A%20a%20Taxonomy%20of%20Deep%20Learning%20Approaches%20for%20Object%20Counting)**\n\n\n> **TL;DR:** We provide a taxonomy that reflects both: the method characteristic as well as the counting problem characteristics in order to derive a systematization of the scientific field. This taxonomy is applied on four case studies.\n\nMany tasks, like process fault detection, disease diagnostic or maintaining security in public places are connected to the process of counting objects from observing image data like camera footage or microscopic images. However, the pivotal process of counting objects based on image data is subject to a lot of error sources when done manually. Therefore, a lot of approaches exist that aim to aid the automation of the counting process. Since most of those methods are of black box nature and have to be applied to real world scenarios, we provide a taxonomy that reflects both: the method characteristic as well as the counting problem characteristics in order to derive a systematization of the scientific field and to provide a tool that acts as a set of guidelines for choosing appropriate counting methods for different situations. We first conduct a literature review, where counting problem characteristics and solutions are extracted and later transformed into a taxonomy. We finally showcase the taxonomy using four case studies that are based on publicly available datasets for the sake of scientific comprehension.\n_____\n\n_____\n**[Crowdsourcing annotation system of object counting dataset for deep learning algorithm](https://doi.org/10.1088/1755-1315/195/1/012063)**\n\nDeep Learning is currently the state-of-the-art technique for various Computer Vision tasks, including object counting. Despite of its high performance, Deep Learning requires a gigantic amount of training data to show its best result. Getting this massive data in reasonable time requires a proper strategy such as crowdsourcing. However, in case of object counting, we found no crowdsourcing system able to effectively collect necessary data. To tackle this problem, we develop a crowdsourcing system to annotate image for object counting dataset. This system is also equipped with validation system to ensure the quality of collected dataset.\n_____\n\n_____\n**[Automatic Pest Counting from Pheromone Trap Images Using Deep Learning Object Detectors for Matsucoccus thunbergianae Monitoring](https://doi.org/10.3390/insects12040342)**\n\nSimple Summary The black pine bast scale, Matsucoccus thunbergianae, is a forest pest that causes widespread damage to black pine; therefore, monitoring this pest is necessary to minimize environmental and economic losses in forests. However, monitoring insects in pheromone traps performed by humans is labor intensive and time consuming. To develop an automated monitoring system, we aimed to develop algorithms that detect and count M. thunbergianae from images of pheromone traps using deep-learning-based object detection algorithms. Object detection models based on deep learning neural networks under various conditions were trained, and the performances of detection and counting were compared and evaluated. In addition, the models were trained to detect small objects well by cropping images into multiple windows. As a result, the algorithms based on deep learning neural networks successfully detected and counted M. thunbergianae. These results showed that accurate and constant pest monitoring is possible using the artificial-intelligence-based methods we proposed. Abstract The black pine bast scale, M. thunbergianae, is a major insect pest of black pine and causes serious environmental and economic losses in forests. Therefore, it is essential to monitor the occurrence and population of M. thunbergianae, and a monitoring method using a pheromone trap is commonly employed. Because the counting of insects performed by humans in these pheromone traps is labor intensive and time consuming, this study proposes automated deep learning counting algorithms using pheromone trap images. The pheromone traps collected in the field were photographed in the laboratory, and the images were used for training, validation, and testing of the detection models. In addition, the image cropping method was applied for the successful detection of small objects in the image, considering the small size of M. thunbergianae in trap images. The detection and counting performance were evaluated and compared for a total of 16 models under eight model conditions and two cropping conditions, and a counting accuracy of 95% or more was shown in most models. This result shows that the artificial intelligence-based pest counting method proposed in this study is suitable for constant and accurate monitoring of insect pests.\n_____\n\n_____\n**[Counting Dense Leaves under Natural Environments via an Improved Deep-Learning-Based Object Detection Algorithm](https://doi.org/10.3390/agriculture11101003)**\n\nThe leaf is the organ that is crucial for photosynthesis and the production of nutrients in plants; as such, the number of leaves is one of the key indicators with which to describe the development and growth of a canopy. The irregular shape and distribution of the blades, as well as the effect of natural light, make the segmentation and detection process of the blades difficult. The inaccurate acquisition of plant phenotypic parameters may affect the subsequent judgment of crop growth status and crop yield. To address the challenge in counting dense and overlapped plant leaves under natural environments, we proposed an improved deep-learning-based object detection algorithm by merging a space-to-depth module, a Convolutional Block Attention Module (CBAM) and Atrous Spatial Pyramid Pooling (ASPP) into the network, and applying the smoothL1 function to improve the loss function of object prediction. We evaluated our method on images of five different plant species collected under indoor and outdoor environments. The experimental results demonstrated that our algorithm which counts dense leaves improved average detection accuracy of 85% to 96%. Our algorithm also showed better performance in both detection accuracy and time consumption compared to other state-of-the-art object detection algorithms.\n_____\n\n_____\n**[Plant Counting of Cotton from UAS Imagery Using Deep Learning-Based Object Detection Framework](https://doi.org/10.3390/rs12182981)**\n\nAssessing plant population of cotton is important to make replanting decisions in low plant density areas, prone to yielding penalties. Since the measurement of plant population in the field is labor intensive and subject to error, in this study, a new approach of image-based plant counting is proposed, using unmanned aircraft systems (UAS; DJI Mavic 2 Pro, Shenzhen, China) data. The previously developed image-based techniques required a priori information of geometry or statistical characteristics of plant canopy features, while also limiting the versatility of the methods in variable field conditions. In this regard, a deep learning-based plant counting algorithm was proposed to reduce the number of input variables, and to remove requirements for acquiring geometric or statistical information. The object detection model named You Only Look Once version 3 (YOLOv3) and photogrammetry were utilized to separate, locate, and count cotton plants in the seedling stage. The proposed algorithm was tested with four different UAS datasets, containing variability in plant size, overall illumination, and background brightness. Root mean square error (RMSE) and R2 values of the optimal plant count results ranged from 0.50 to 0.60 plants per linear meter of row (number of plants within 1 m distance along the planting row direction) and 0.96 to 0.97, respectively. The object detection algorithm, trained with variable plant size, ground wetness, and lighting conditions generally resulted in a lower detection error, unless an observable difference of developmental stages of cotton existed. The proposed plant counting algorithm performed well with 0–14 plants per linear meter of row, when cotton plants are generally separable in the seedling stage. This study is expected to provide an automated methodology for in situ evaluation of plant emergence using UAS data.\n_____\n\n_____\n**[Vehicle Counting System using Deep Learning and Multi-Object Tracking Methods](https://doi.org/10.1177/0361198120912742)**\n\nUsing deep learning technology and multi-object tracking method to count vehicles accurately in different traffic conditions is a hot research topic in the field of intelligent transportation. In this paper, first, a vehicle dataset from the perspective of highway surveillance cameras is constructed, and the vehicle detection model is obtained by training using the You Only Look Once (YOLO) version 3 network. Second, an improved multi-scale and multi-feature tracking algorithm based on a kernel correlation filter (KCF) algorithm is proposed to avoid the KCF extracting single features and single-scale defects. Combining the intersection over union (IoU) similarity measure and the row-column optimal association criterion proposed in this paper, matching strategy is used to process the vehicles that are not detected and wrongly detected, thereby obtaining complete vehicle trajectories. Finally, according to the trajectory of the vehicle, the traveling direction of the vehicle is automatically determined, and the setting position of the detecting line is automatically updated to obtain the vehicle count result accurately. Experiments were conducted in a variety of traffic scenes and compared with published data. The experimental results show that the proposed method achieves high accuracy of vehicle detection while maintaining accuracy and precision in tracking multiple objects, and obtains accurate vehicle counting results which can meet real-time processing requirements. The algorithm presented in this paper has practical application for vehicle counting in complex highway scenes.\n_____\n\n_____\n**[Occlusion Robust Wheat Ear Counting Algorithm Based on Deep Learning](https://doi.org/10.3389/fpls.2021.645899)**\n\nCounting the number of wheat ears in images under natural light is an important way to evaluate the crop yield, thus, it is of great significance to modern intelligent agriculture. However, the distribution of wheat ears is dense, so the occlusion and overlap problem appears in almost every wheat image. It is difficult for traditional image processing methods to solve occlusion problem due to the deficiency of high-level semantic features, while existing deep learning based counting methods did not solve the occlusion efficiently. This article proposes an improved EfficientDet-D0 object detection model for wheat ear counting, and focuses on solving occlusion. First, the transfer learning method is employed in the pre-training of the model backbone network to extract the high-level semantic features of wheat ears. Secondly, an image augmentation method Random-Cutout is proposed, in which some rectangles are selected and erased according to the number and size of the wheat ears in the images to simulate occlusion in real wheat images. Finally, convolutional block attention module (CBAM) is adopted into the EfficientDet-D0 model after the backbone, which makes the model refine the features, pay more attention to the wheat ears and suppress other useless background information. Extensive experiments are done by feeding the features to detection layer, showing that the counting accuracy of the improved EfficientDet-D0 model reaches 94%, which is about 2% higher than the original model, and false detection rate is 5.8%, which is the lowest among comparative methods.\n_____\n\n_____\n**[Applications of object detection in modular construction based on a comparative evaluation of deep learning algorithms](https://doi.org/10.1108/CI-02-2020-0017)**\n\n\nPurpose\nThe practice of artificial intelligence (AI) is increasingly being promoted by technology developers. However, its adoption rate is still reported as low in the construction industry due to a lack of expertise and the limited reliable applications for AI technology. Hence, this paper aims to present the detailed outcome of experimentations evaluating the applicability and the performance of AI object detection algorithms for construction modular object detection.\n\n\nDesign/methodology/approach\nThis paper provides a thorough evaluation of two deep learning algorithms for object detection, including the faster region-based convolutional neural network (faster RCNN) and single shot multi-box detector (SSD). Two types of metrics are also presented; first, the average recall and mean average precision by image pixels; second, the recall and precision by counting. To conduct the experiments using the selected algorithms, four infrastructure and building construction sites are chosen to collect the required data, including a total of 990 images of three different but common modular objects, including modular panels, safety barricades and site fences.\n\n\nFindings\nThe results of the comprehensive evaluation of the algorithms show that the performance of faster RCNN and SSD depends on the context that detection occurs. Indeed, surrounding objects and the backgrounds of the objects affect the level of accuracy obtained from the AI analysis and may particularly effect precision and recall. The analysis of loss lines shows that the loss lines for selected objects depend on both their geometry and the image background. The results on selected objects show that faster RCNN offers higher accuracy than SSD for detection of selected objects.\n\n\nResearch limitations/implications\nThe results show that modular object detection is crucial in construction for the achievement of the required information for project quality and safety objectives. The detection process can significantly improve monitoring object installation progress in an accurate and machine-based manner avoiding human errors. The results of this paper are limited to three construction sites, but future investigations can cover more tasks or objects from different construction sites in a fully automated manner.\n\n\nOriginality/value\nThis paper’s originality lies in offering new AI applications in modular construction, using a large first-hand data set collected from three construction sites. Furthermore, the paper presents the scientific evaluation results of implementing recent object detection algorithms across a set of extended metrics using the original training and validation data sets to improve the generalisability of the experimentation. This paper also provides the practitioners and scholars with a workflow on AI applications in the modular context and the first-hand referencing data.\n\n_____\n\n_____\n**[Leaf Counting in Rice (Oryza Sativa L.) Using Object Detection: A Deep Learning Approach](https://doi.org/10.1109/IGARSS39084.2020.9324153)**\n\nLeaf count is one of the crucial tasks in plant phenotyping, and leaves are the basic unit of plant architecture involved in photosynthesis, growth, and yield of a plant. Therefore, the total number of leaves per plant is considered as one of the essential physio-morphological plant traits for phenotyping. The current work proposes to estimate the total number of leaves of a rice plant by detecting their leaves tips. A rice plant has a single tip for a single leaf. Hence, this proposed framework counts the total number of leaves by counting the number of leaves tips equal to the number of leaves. You Only Look Once (YOLO) algorithm is used for the detection of the leaves tips as an object. This hypothesis builds a basis for counting the total number of leaves in a plant like rice, and similar field crops such as wheat (Triticum aestivum L), maize (Zea mays L.), sorghum (Sorghum bicolor), barley (Hordeum vulgare L.). The model detected leaves of a rice plant (RGB images) by detecting corresponding leaves tips with YOLO having average accuracy up to 82% and IOU around 0.53-0.60 and estimates the number of leaves in a plant by counting predicted bounding boxes around tips. The model also performed well with the wheat crop.\n_____\n\n_____\n**[Central Object Segmentation by Deep Learning to Continuously Monitor Fruit Growth through RGB Images](https://doi.org/10.3390/s21216999)**\n\nMonitoring fruit growth is useful when estimating final yields in advance and predicting optimum harvest times. However, observing fruit all day at the farm via RGB images is not an easy task because the light conditions are constantly changing. In this paper, we present CROP (Central Roundish Object Painter). The method involves image segmentation by deep learning, and the architecture of the neural network is a deeper version of U-Net. CROP identifies different types of central roundish fruit in an RGB image in varied light conditions, and creates a corresponding mask. Counting the mask pixels gives the relative two-dimensional size of the fruit, and in this way, time-series images may provide a non-contact means of automatically monitoring fruit growth. Although our measurement unit is different from the traditional one (length), we believe that shape identification potentially provides more information. Interestingly, CROP can have a more general use, working even for some other roundish objects. For this reason, we hope that CROP and our methodology yield big data to promote scientific advancements in horticultural science and other fields.\n_____\n\n_____\n**[A NOVEL DEEP LEARNING BASED METHOD FOR DETECTION AND COUNTING OF VEHICLES IN URBAN TRAFFIC SURVEILLANCE SYSTEMS](https://doi.org/10.5194/isprs-archives-xliii-b2-2021-793-2021)**\n\nIn intelligent transportation systems (ITS), it is essential to obtain reliable statistics of the vehicular flow in order to create urban traffic management strategies. These systems have benefited from the increase in computational resources and the improvement of image processing methods, especially in object detection based on deep learning. This paper proposes a method for vehicle counting composed of three stages: object detection, tracking and trajectory processing. In order to select the detection model with the best trade-off between accuracy and speed, the following one-stage detection models were compared: SSD512, CenterNet, Efficiedet-D0 and YOLO family models (v2, v3 and v4). Experimental results conducted on the benchmark dataset show that the best rates among the detection models were obtained using YOLOv4 with mAP =87% and a processing speed of 18 FPS. On the other hand, the accuracy obtained in the proposed counting method was 94% with a real-time processing rate lower than 1.9.\n_____\n\n_____\n**[YOLO-Based Deep Learning Framework for Olive Fruit Fly Detection and Counting](https://doi.org/10.1109/ACCESS.2021.3088075)**\n\nThe olive fruit fly can damage up to 100% of the harvested fruit and can cause up to 80% reduction of the value of the resulting olive oil. Therefore, it is important to early detect its presence in the olive orchard to take the appropriate chemical or biological countermeasures as early as possible. Traps filled with attractant pheromones are typically deployed across the orchard to attract and capture the flies. Traditionally, the captured flies were manually counted which is error prone. Recently, the traps are employed with cameras and communication devices to send pictures of the captured flies to experts for analysis which is also error prone and inefficient. Consequently, machine and deep learning have been exploited to develop fully automated and accurate detection that does not include human in the loop. Such a learning problem is challenging due to the small size of the detected object, the differences in the light conditions at which pictures were taken, and the lack of enough data to train the learning model. In this paper, we present a deep learning framework for detecting and counting the number of olive fruit flies that exploits data augmentation to increase the dataset size, includes negative samples in the training to improve the detection accuracy, and normalizes the images to the color of the trap background, i.e., yellow, to unify the illumination conditions. The results of the proposed framework show a precision of 0.84, a recall of 0.97, an F1-score of 0.9 and mean Average Precision (mAP) of 96.68% which significantly outperforms existing pest detection systems.\n_____\n\n_____\n**[A Deep Learning Bidirectional Temporal Tracking Algorithm for Automated Blood Cell Counting from Non-invasive Capillaroscopy Videos](https://doi.org/10.1007/978-3-030-87237-3_40)**\n\n\n> **TL;DR:** We propose a deep learning multi-cell tracking model, CycleTrack, which achieves accurate blood cell counting from capillaroscopic videos. Our approach combines two simple online tracking models, SORT and CenterTrack, and is tailored to features of capillary blood cell flow. Blood cells are tracked by displacement vectors in two opposing temporal directions (forward- and backward-tracking) between consecutive frames. The proposed model outperforms other baseline trackers, achieving 65.57% Multiple Object Tracking Accuracy and 73.95% ID F1 score on test videos. Compared to manual blood cell counting, CycleTrack achieves 96.58 $\\pm$\n\nOblique back-illumination capillaroscopy has recently been introduced as a method for high-quality, non-invasive blood cell imaging in human capillaries. To make this technique practical for clinical blood cell counting, solutions for automatic processing of acquired videos are needed. Here, we take the first step towards this goal, by introducing a deep learning multi-cell tracking model, named CycleTrack, which achieves accurate blood cell counting from capillaroscopic videos. CycleTrack combines two simple online tracking models, SORT and CenterTrack, and is tailored to features of capillary blood cell flow. Blood cells are tracked by displacement vectors in two opposing temporal directions (forward- and backward-tracking) between consecutive frames. This approach yields accurate tracking despite rapidly moving and deforming blood cells. The proposed model outperforms other baseline trackers, achieving 65.57% Multiple Object Tracking Accuracy and 73.95% ID F1 score on test videos. Compared to manual blood cell counting, CycleTrack achieves 96.58 $\\pm$ 2.43% cell counting accuracy among 8 test videos with 1000 frames each compared to 93.45% and 77.02% accuracy for independent CenterTrack and SORT almost without additional time expense. It takes 800s to track and count approximately 8000 blood cells from 9,600 frames captured in a typical one-minute video. Moreover, the blood cell velocity measured by CycleTrack demonstrates a consistent, pulsatile pattern within the physiological range of heart rate. Lastly, we discuss future improvements for the CycleTrack framework, which would enable clinical translation of the oblique back-illumination microscope towards a real-time and non-invasive point-of-care blood cell counting and analyzing technology.\n_____\n\n_____\n**[A NOVEL DEEP LEARNING BASED METHOD FOR DETECTION AND COUNTING OF VEHICLES IN URBAN TRAFFIC SURVEILLANCE SYSTEMS](https://doi.org/10.5194/isprs-archives-xliii-b2-2021-793-2021)**\n\n\n> **TL;DR:** This paper proposes a method for vehicle counting composed of three stages: object detection, tracking and trajectory processing. In order to select the detection model with the best trade-off between accuracy and speed, the following one-stage detection models were compared: SSD512, CenterNet, Efficiedet-D0 and YOLO family models (v2, v3 and v4). Experimental results conducted on the benchmark dataset show that the best rates among the detection models were obtained using YOLOv4 with mAP =87% and a processing speed of 18 FPS. On the other hand, the accuracy obtained in the\n\nIn intelligent transportation systems (ITS), it is essential to obtain reliable statistics of the vehicular flow in order to create urban traffic management strategies. These systems have benefited from the increase in computational resources and the improvement of image processing methods, especially in object detection based on deep learning. This paper proposes a method for vehicle counting composed of three stages: object detection, tracking and trajectory processing. In order to select the detection model with the best trade-off between accuracy and speed, the following one-stage detection models were compared: SSD512, CenterNet, Efficiedet-D0 and YOLO family models (v2, v3 and v4). Experimental results conducted on the benchmark dataset show that the best rates among the detection models were obtained using YOLOv4 with mAP =87% and a processing speed of 18 FPS. On the other hand, the accuracy obtained in the proposed counting method was 94% with a real-time processing rate lower than 1.9.\n_____\n\n_____\n**[Multiple Object Tracking using Deep Learning with YOLO V5](https://scholar.google.com/scholar?q=Multiple%20Object%20Tracking%20using%20Deep%20Learning%20with%20YOLO%20V5)**\n\nThe MOT (Multiple Object Tracking) is an important tool in the modern world. It has various uses like object detection, counting objects, security tools ,etc. The Object tracking is a prominent technology in image processing which has a large future scope. The MOT has made significant growth in a few years due to deep learning, computer vision, machine learning, etc. This paper aims to provide a software solution that keeps track of the objects so that it can handle object list and count. By using YOLO “You Only Look Once” Technology with the help of Pytorch, the system aims in object detection, tracking and counting. Also unlike the general yolo object detection tool which detects all objects at the same time ,this MOT system also detects only objects which are needed to be detected by the user and thus helps in improving the performance of the system.\n_____\n\n_____\n**[Tomato Fruit Detection and Counting in Greenhouses Using Deep Learning](https://doi.org/10.3389/fpls.2020.571299)**\n\nAccurately detecting and counting fruits during plant growth using imaging and computer vision is of importance not only from the point of view of reducing labor intensive manual measurements of phenotypic information, but also because it is a critical step toward automating processes such as harvesting. Deep learning based methods have emerged as the state-of-the-art techniques in many problems in image segmentation and classification, and have a lot of promise in challenging domains such as agriculture, where they can deal with the large variability in data better than classical computer vision methods. This paper reports results on the detection of tomatoes in images taken in a greenhouse, using the MaskRCNN algorithm, which detects objects and also the pixels corresponding to each object. Our experimental results on the detection of tomatoes from images taken in greenhouses using a RealSense camera are comparable to or better than the metrics reported by earlier work, even though those were obtained in laboratory conditions or using higher resolution images. Our results also show that MaskRCNN can implicitly learn object depth, which is necessary for background elimination.\n_____\n\n_____\n**[Towards a low-cost embedded vehicle counting system based on deep-learning for traffic management applications](https://doi.org/10.1109/CHILECON54041.2021.9702914)**\n\n\n> **TL;DR:** The Kalman filter is more efficient than the centroid tracking algorithm, and it produces fewer errors.\n\nThis paper explores the feasibility of using a low-cost embedded system for real-time vehicle detection and counting through the use of deep neural networks. It compares the performance of two different object tracking methods, the Kalman filter with the Hungarian algorithm and the centroid tracking algorithm. The experimentation proved that the efficiency of the implemented algorithms was above the 92% and 98% for the centroid tracking algorithm and Kalman filter with the Hungarian algorithm, respectively. Also, the Kalman filter produced fewer errors overcoming the centroid tracking algorithm.\n_____\n\n_____\n**[Recognizing and counting Dendrocephalus brasiliensis (Crustacea: Anostraca) cysts using deep learning](https://doi.org/10.1371/journal.pone.0248574)**\n\n\n> **TL;DR:** We propose an automatized approach for the detection and counting of Dendrocephalus brasiliensis cysts from images captured by a digital microscope. For this purpose, we built the DBrasiliensis dataset, a repository with 246 images containing 5141 cysts of Dendrocephalus brasiliensis. Then, we trained two state-of-the-art object detection methods, YOLOv3 (You Only Look Once) and Faster R-CNN (Region-based Convolutional Neural Networks), on DBrasiliensis dataset in order to compare them under both cyst detection and counting tasks.\n\nThe Dendrocephalus brasiliensis, a native species from South America, is a freshwater crustacean well explored in conservational and productive activities. Its main characteristics are its rusticity and resistance cysts production, in which the hatching requires a period of dehydration. Independent of the species utilization nature, it is essential to manipulate its cysts, such as the counting using microscopes. Manually counting is a difficult task, prone to errors, and that also very time-consuming. In this paper, we propose an automatized approach for the detection and counting of Dendrocephalus brasiliensis cysts from images captured by a digital microscope. For this purpose, we built the DBrasiliensis dataset, a repository with 246 images containing 5141 cysts of Dendrocephalus brasiliensis. Then, we trained two state-of-the-art object detection methods, YOLOv3 (You Only Look Once) and Faster R-CNN (Region-based Convolutional Neural Networks), on DBrasiliensis dataset in order to compare them under both cyst detection and counting tasks. Experiments showed evidence that YOLOv3 is superior to Faster R-CNN, achieving an accuracy rate of 83,74%, R2 of 0.88, RMSE (Root Mean Square Error) of 3.49, and MAE (Mean Absolute Error) of 2.24 on cyst detection and counting. Moreover, we showed that is possible to infer the number of cysts of a substrate, with known weight, by performing the automated counting of some of its samples. In conclusion, the proposed approach using YOLOv3 is adequate to detect and count Dendrocephalus brasiliensis cysts. The DBrasiliensis dataset can be accessed at: https://doi.org/10.6084/m9.figshare.13073240.\n_____\n\n_____\n**[SpikeSegNet-a deep learning approach utilizing encoder-decoder network with hourglass for spike segmentation and counting in wheat plant from visual imaging](https://doi.org/10.1186/s13007-020-00582-9)**\n\nBackground High throughput non-destructive phenotyping is emerging as a significant approach for phenotyping germplasm and breeding populations for the identification of superior donors, elite lines, and QTLs. Detection and counting of spikes, the grain bearing organs of wheat, is critical for phenomics of a large set of germplasm and breeding lines in controlled and field conditions. It is also required for precision agriculture where the application of nitrogen, water, and other inputs at this critical stage is necessary. Further, counting of spikes is an important measure to determine yield. Digital image analysis and machine learning techniques play an essential role in non-destructive plant phenotyping analysis. Results In this study, an approach based on computer vision, particularly object detection, to recognize and count the number of spikes of the wheat plant from the digital images is proposed. For spike identification, a novel deep-learning network, SpikeSegNet, has been developed by combining two proposed feature networks: Local Patch extraction Network (LPNet) and Global Mask refinement Network (GMRNet). In LPNet, the contextual and spatial features are learned at the local patch level. The output of LPNet is a segmented mask image, which is further refined at the global level using GMRNet. Visual (RGB) images of 200 wheat plants were captured using LemnaTec imaging system installed at Nanaji Deshmukh Plant Phenomics Centre, ICAR-IARI, New Delhi. The precision, accuracy, and robustness (F 1 score) of the proposed approach for spike segmentation are found to be 99.93%, 99.91%, and 99.91%, respectively. For counting the number of spikes, “analyse particles”—function of imageJ was applied on the output image of the proposed SpikeSegNet model. For spike counting, the average precision, accuracy, and robustness are 99%, 95%, and 97%, respectively. SpikeSegNet approach is tested for robustness with illuminated image dataset, and no significant difference is observed in the segmentation performance. Conclusion In this study, a new approach called as SpikeSegNet has been proposed based on combined digital image analysis and deep learning techniques. A dedicated deep learning approach has been developed to identify and count spikes in the wheat plants. The performance of the approach demonstrates that SpikeSegNet is an effective and robust approach for spike detection and counting. As detection and counting of wheat spikes are closely related to the crop yield, and the proposed approach is also non-destructive, it is a significant step forward in the area of non-destructive and high-throughput phenotyping of wheat.\n_____\n\n_____\n**[A Simple Vehicle Counting System Using Deep Learning with YOLOv3 Model](https://doi.org/10.29207/resti.v4i3.1871)**\n\nDeep Learning is a popular Machine Learning algorithm that is widely used in many areas in current daily life. Its robust performance and ready-to-use frameworks and architectures enables many people to develop various Deep Learning-based software or systems to support human tasks and activities. Traffic monitoring is one area that utilizes Deep Learning for several purposes. By using cameras installed in some spots on the roads, many tasks such as vehicle counting, vehicle identification, traffic violation monitoring, vehicle speed monitoring, etc. can be realized. In this paper, we discuss a Deep Learning implementation to create a vehicle counting system without having to track the vehicles movements. To enhance the system performance and to reduce time in deploying Deep Learning architecture, hence pretrained model of YOLOv3 is used in this research due to its good performance and moderate computational time in object detection. This research aims to create a simple vehicle counting system to help human in classify and counting the vehicles that cross the street. The counting is based on four types of vehicle, i.e. car, motorcycle, bus, and truck, while previous research counts the car only. As the result, our proposed system capable to count the vehicles crossing the road based on video captured by camera with the highest accuracy of 97.72%.\n_____\n\n_____\n**[Accurate stacked-sheet counting method based on deep learning.](https://doi.org/10.1364/josaa.387390)**\n\n\n> **TL;DR:** This paper proposes a new stacked-sheet counting method with a deep learning approach using the U-Net architecture. A specific dataset according to the characteristics of stack side images is collected. The stripe of the center line of each sheet is used for semantic segmentation, and the complete side images of the slices are segmented via training with small image patches and testing with original large images. With this model, each pixel is classified by multi-layer convolution and deconvolution to determine whether it is the target object to be detected. After the model is trained, the test set is used to test the model, and a center region\n\nThe accurate counting of laminated sheets, such as packing or printing sheets in industry, is extremely important because it greatly affects the economic cost. However, the different thicknesses, adhesion properties, and breakage points and the low contrast of sheets remain challenges to traditional counting methods based on image processing. This paper proposes a new stacked-sheet counting method with a deep learning approach using the U-Net architecture. A specific dataset according to the characteristics of stack side images is collected. The stripe of the center line of each sheet is used for semantic segmentation, and the complete side images of the slices are segmented via training with small image patches and testing with original large images. With this model, each pixel is classified by multi-layer convolution and deconvolution to determine whether it is the target object to be detected. After the model is trained, the test set is used to test the model, and a center region segmentation map based on the pixel points is obtained. By calculating the statistical median value of centerline points across different sections in these segmented images, the number of sheets can be obtained. Compared with traditional image algorithms in real product counting experiments, the proposed method can achieve better performance with higher accuracy and a lower error rate.\n_____\n\n_____\n**[Automatic Detection and Counting of Lymphocytes from Immunohistochemistry Cancer Images Using Deep Learning](https://doi.org/10.1007/s40846-020-00545-4)**\n\nPurpose Cancer is one of the most life-threatening and devastating diseases in the world. The generally recognized standard for cancer staging is the TNM staging system. In addition, a new parameter called the Immunoscore has been developed recently to assess the survival rate of patients. The Immunoscore is based on counts of CD3+ and CD8+ lymphocytes in the tumour core and margin. Counting the number of lymphocytes is a tedious process for pathologists. This paper examines the use of deep learning techniques for automatic detection and counting of lymphocytes from immunohistochemistry images of breast, colon, and prostate cancers. Methods We used an object detector called Faster R-CNN with four feature extractors: Resnet-50, VGG-16, Inception-V2, and Resnet-101 for automatic detection and counting of lymphocytes. A total of 11,136 lymphocytes were annotated after performing data augmentation on 1228 images. The test images are separated into three regions of interest (ROI): scattered lymphocytes, groups of lymphocytes, and artefacts. In each ROI, the performance of the object detector was checked by evaluation metrics. Results On comparing the F1-score for all three ROIs, we found that Resnet-101 provides better performance than the other feature extractors when using Faster R-CNN. The mean error in lymphocyte count for all ROIs appears to be insignificant. The detection time for a single image is less for VGG-16 than for other feature extractors. Conclusion This study presents a fine-tuned Faster R-CNN object detector for automatic detection and counting of lymphocytes in three different cancer tissues for scoring lymphocytes. Our results suggest that the Faster R-CNN method is efficient and yields good results. Thus, the proposed method can assist pathologists in providing a cancer prognosis.\n_____\n" } }