Hi everyone ! Since this competition seems fairly similar to the Cornell Birdcall Identification, I'll just drop here the resources that I had put up together at the time. Feel free to check our solution to this competition in the great post by @theoviel here.
\nFirst of all, on Kaggle, you have a lot of interesting notebooks, competitions or datasets to get started :
\nMore generally, on sound classification using deep learning :
\nMore specifically, on bird sound classification :
\nIf you have, like me, a special interest to Generative Adversarial Networks & DeepGANs:
\n… my conclusion ? Well, CNN seems to be a good place to start (at least, very widely used in the literature). You have sample codes from a lot of papers available online, for example here for the paper Large-Scale Bird Sound Classification using\nConvolutional Neural Networks. But don't get stuck at reproducing peer-reviewed literature ! Try new things, and happy kaggling !
", "messages": [ { "id": 1259901, "postDate": "2021-04-01T18:16:13.420Z", "content": "Hi everyone ! Since this competition seems fairly similar to the Cornell Birdcall Identification, I'll just drop here the resources that I had put up together at the time. Feel free to check our solution to this competition in the great post by @theoviel here.
\nFirst of all, on Kaggle, you have a lot of interesting notebooks, competitions or datasets to get started :
\nMore generally, on sound classification using deep learning :
\nMore specifically, on bird sound classification :
\nIf you have, like me, a special interest to Generative Adversarial Networks & DeepGANs:
\n… my conclusion ? Well, CNN seems to be a good place to start (at least, very widely used in the literature). You have sample codes from a lot of papers available online, for example here for the paper Large-Scale Bird Sound Classification using\nConvolutional Neural Networks. But don't get stuck at reproducing peer-reviewed literature ! Try new things, and happy kaggling !
", "rawMarkdown": "Hi everyone ! Since this competition seems fairly similar to the [Cornell Birdcall Identification](https://www.kaggle.com/c/birdsong-recognition/overview), I'll just drop here the resources that I had put up together at the time. Feel free to check our solution to this competition in the great post by @theoviel [here](https://www.kaggle.com/c/birdsong-recognition/discussion/183199).\n\nFirst of all, on Kaggle, you have a lot of interesting notebooks, competitions or datasets to get started : \n- https://www.kaggle.com/mmoreaux/environmental-sound-classification-50\n- https://www.kaggle.com/c/freesound-audio-tagging/data\n- https://www.kaggle.com/devilsknight/sound-classification-using-spectrogram-images\n- https://www.kaggle.com/mychen76/automatic-urban-sound-classification-with-cnn\n- https://www.kaggle.com/ashishpatel26/feature-extraction-from-audio\n- https://www.kaggle.com/fizzbuzz/beginner-s-guide-to-audio-data\n\nMore generally, on **sound classification using deep learning** : \n- a very complete [review of literature](https://www.mdpi.com/2076-3417/10/6/2020/pdf)\n- [The basics, on medium](https://medium.com/@mikesmales/sound-classification-using-deep-learning-8bc2aa1990b7)\n- a wide library of available papers [here](https://paperswithcode.com/task/audio-classification)\n- [if you like tensorflow\n](https://www.iotforall.com/tensorflow-sound-classification-machine-learning-applications/)\n- ... or if you're more of a torch person : [here](https://pytorch.org/tutorials/beginner/audio_classifier_tutorial.html?highlight=audio) and [here](https://medium.com/@hasithsura/audio-classification-d37a82d6715)\n- an interesting idea with [Deep Generative Architectures](Deep Generative Architectures)\n\nMore specifically, on **bird sound classification** : \n- [Women in machine learning and Data Science](https://towardsdatascience.com/sound-based-bird-classification-965d0ecacb2b) project (girls rule 🙌 )\n- on CEUR you have 2 different works with CNN : [one from 2017](http://ceur-ws.org/Vol-1866/paper_143.pdf) and [one from 2019](http://ceur-ws.org/Vol-2380/paper_68.pdf)\n- using more handcrafted features, but still CNN : [this one on australian birds](https://www.sciencedirect.com/science/article/abs/pii/S1574954118302991)\n- [kernel-based machine learning](https://asa.scitation.org/doi/10.1121/1.5004570)\n- still CNN, on researchGate, you have [this one](https://www.researchgate.net/publication/328836649_Bird_Sound_Recognition_Using_a_Convolutional_Neural_Network) and [this one](https://www.researchgate.net/publication/322144806_Large-Scale_Bird_Sound_Classification_using_Convolutional_Neural_Networks)\n- if you feel allergic to handcrafted features or magic, you have [this paper on arxiv\n](https://arxiv.org/pdf/1807.05812.pdf)\n- [that very innovative one](https://arxiv.org/ftp/arxiv/papers/1803/1803.01107.pdf) using transfer learning, still on arxiv\n\nIf you have, like me, a special interest to Generative Adversarial Networks & DeepGANs:\n- [towardsdatascience](https://towardsdatascience.com/synthesizing-audio-with-generative-adversarial-networks-8e0308184edd), of course\n- for data augmentation : [environmental sounds classification](https://ieeexplore.ieee.org/document/8902819)\n- [masked CNN](https://www.sciencedirect.com/science/article/pii/S1568494620300132) (bit out of topic, but didn't know where to put it)\n- [audio generation with GAN](https://medium.com/neuronio/audio-generation-with-gans-428bc2de5a89)\n- [Unsupervised Feature Learning for Environmental Sound Classification Using Cycle Consistent Generative Adversarial Network](https://www.groundai.com/project/unsupervised-feature-learning-for-environmental-sound-classification-using-cycle-consistent-generative-adversarial-network/1)\n- still [data augmentation with GAN\n](https://www.researchgate.net/publication/335739623_Data_Augmentation_Using_Generative_Adversarial_Network_for_Environmental_Sound_Classification)\n\n... my conclusion ? Well, CNN seems to be a good place to start (at least, very widely used in the literature). You have sample codes from a lot of papers available online, for example [here](https://github.com/kahst/BirdCLEF2017) for the paper [Large-Scale Bird Sound Classification using\nConvolutional Neural Networks](http://ceur-ws.org/Vol-1866/paper_143.pdf). But don't get stuck at reproducing peer-reviewed literature ! Try new things, and happy kaggling !", "votes": 18 } ], "comments": [], "raw_markdown_by_id": { "1259901": "Hi everyone ! Since this competition seems fairly similar to the [Cornell Birdcall Identification](https://www.kaggle.com/c/birdsong-recognition/overview), I'll just drop here the resources that I had put up together at the time. Feel free to check our solution to this competition in the great post by @theoviel [here](https://www.kaggle.com/c/birdsong-recognition/discussion/183199).\n\nFirst of all, on Kaggle, you have a lot of interesting notebooks, competitions or datasets to get started : \n- https://www.kaggle.com/mmoreaux/environmental-sound-classification-50\n- https://www.kaggle.com/c/freesound-audio-tagging/data\n- https://www.kaggle.com/devilsknight/sound-classification-using-spectrogram-images\n- https://www.kaggle.com/mychen76/automatic-urban-sound-classification-with-cnn\n- https://www.kaggle.com/ashishpatel26/feature-extraction-from-audio\n- https://www.kaggle.com/fizzbuzz/beginner-s-guide-to-audio-data\n\nMore generally, on **sound classification using deep learning** : \n- a very complete [review of literature](https://www.mdpi.com/2076-3417/10/6/2020/pdf)\n- [The basics, on medium](https://medium.com/@mikesmales/sound-classification-using-deep-learning-8bc2aa1990b7)\n- a wide library of available papers [here](https://paperswithcode.com/task/audio-classification)\n- [if you like tensorflow\n](https://www.iotforall.com/tensorflow-sound-classification-machine-learning-applications/)\n- ... or if you're more of a torch person : [here](https://pytorch.org/tutorials/beginner/audio_classifier_tutorial.html?highlight=audio) and [here](https://medium.com/@hasithsura/audio-classification-d37a82d6715)\n- an interesting idea with [Deep Generative Architectures](Deep Generative Architectures)\n\nMore specifically, on **bird sound classification** : \n- [Women in machine learning and Data Science](https://towardsdatascience.com/sound-based-bird-classification-965d0ecacb2b) project (girls rule 🙌 )\n- on CEUR you have 2 different works with CNN : [one from 2017](http://ceur-ws.org/Vol-1866/paper_143.pdf) and [one from 2019](http://ceur-ws.org/Vol-2380/paper_68.pdf)\n- using more handcrafted features, but still CNN : [this one on australian birds](https://www.sciencedirect.com/science/article/abs/pii/S1574954118302991)\n- [kernel-based machine learning](https://asa.scitation.org/doi/10.1121/1.5004570)\n- still CNN, on researchGate, you have [this one](https://www.researchgate.net/publication/328836649_Bird_Sound_Recognition_Using_a_Convolutional_Neural_Network) and [this one](https://www.researchgate.net/publication/322144806_Large-Scale_Bird_Sound_Classification_using_Convolutional_Neural_Networks)\n- if you feel allergic to handcrafted features or magic, you have [this paper on arxiv\n](https://arxiv.org/pdf/1807.05812.pdf)\n- [that very innovative one](https://arxiv.org/ftp/arxiv/papers/1803/1803.01107.pdf) using transfer learning, still on arxiv\n\nIf you have, like me, a special interest to Generative Adversarial Networks & DeepGANs:\n- [towardsdatascience](https://towardsdatascience.com/synthesizing-audio-with-generative-adversarial-networks-8e0308184edd), of course\n- for data augmentation : [environmental sounds classification](https://ieeexplore.ieee.org/document/8902819)\n- [masked CNN](https://www.sciencedirect.com/science/article/pii/S1568494620300132) (bit out of topic, but didn't know where to put it)\n- [audio generation with GAN](https://medium.com/neuronio/audio-generation-with-gans-428bc2de5a89)\n- [Unsupervised Feature Learning for Environmental Sound Classification Using Cycle Consistent Generative Adversarial Network](https://www.groundai.com/project/unsupervised-feature-learning-for-environmental-sound-classification-using-cycle-consistent-generative-adversarial-network/1)\n- still [data augmentation with GAN\n](https://www.researchgate.net/publication/335739623_Data_Augmentation_Using_Generative_Adversarial_Network_for_Environmental_Sound_Classification)\n\n... my conclusion ? Well, CNN seems to be a good place to start (at least, very widely used in the literature). You have sample codes from a lot of papers available online, for example [here](https://github.com/kahst/BirdCLEF2017) for the paper [Large-Scale Bird Sound Classification using\nConvolutional Neural Networks](http://ceur-ws.org/Vol-1866/paper_143.pdf). But don't get stuck at reproducing peer-reviewed literature ! Try new things, and happy kaggling !" } }