Here is a set of resources I found very useful. Feel free to add other resources you find in the comments and I will add to this resource and mention you. Thanks!
\n\n-
\n\nAudio Based Bird Species Identification using Deep Learning Techniques - This paper was one of the first to use deep learning methods over KNN and tree methods. It also elaborates on some really nice data preparation methods and is a fantastic introduction to what we will be doing here. It also elaborates on feature generation such as time shifting, pitch shifting, and more.
\n\nLink: http://ceur-ws.org/Vol-1609/16090547.pdf
\n\nThesis - Identifying Birds by Sound: Large-scale Acoustic Event Recognition for Avian Activity Monitoring by Stefan Kahl - Probably one of the most thorough resources on this page (A 200+ page thesis). Has an enormous amount of background information and an incredible amount of detail around different specific features and NN arhitectures as they deal with this specific problem.
\n\nLink: https://monarch.qucosa.de/api/qucosa%3A36986/attachment/ATT-0/
\n\nCornell's BirdNet - Cornell has already been reseraching this problem intensively and has deployed several ANN based classifiers online. It may be a good idea to download the app and play around with this to get a feel for what we are trying to do with longer, noisier, and more varied field recordings.\nhttps://birdnet.cornell.edu
\n\n-\nAgnieszka Mikołajczyk - has already put together an amazing set of resources on bird song recognition, many of which are reproduced here as well with more elaboration. - Link: https://github.com/AgaMiko/bird-recognition-review
\n\n-
\n\nAutomatic acoustic detection of birds through deep learning: the first Bird Audio Detection challenge - Used different datasets sourced from birds across Europe. Used Mel Frequency Spectral Coefficients (MFCCs) and Gaussian Mixture Models as a baseline method. Also used
\n\n-
\n\nA Baseline for Large-Scale Bird Species Identification in Field Recordings - Used the BirdCLEF challenge data and a CNN on spectrograms. Used MEL-scale log-amplitude spectrograms.. Trained on ResNet variations and had a goal of applying to mobile devices (and hence tried to use computationally inexpensive techniques).
\n\nLink: http://ceur-ws.org/Vol-2125/paper_85.pdf
\n\n-
\n\nBird Sound Recognition Using a Convolutional Neural Network - Also used Xeno Canto data through pretrained CNN models. Inputs were spectrograms. Limitation: seemed to work best with a low number of predicted classes (bird species).
\n\n\n\n-
\n\nBird Sound Classification using Convolutional Neural Networks - Applied similar process as above to Xeno Canto bird recordings using ResNet and Inception-v3 models.
\nLink: http://ceur-ws.org/Vol-2380/paper_68.pdf
Here is a set of resources I found very useful. Feel free to add other resources you find in the comments and I will add to this resource and mention you. Thanks!
\n\n-
\n\nAudio Based Bird Species Identification using Deep Learning Techniques - This paper was one of the first to use deep learning methods over KNN and tree methods. It also elaborates on some really nice data preparation methods and is a fantastic introduction to what we will be doing here. It also elaborates on feature generation such as time shifting, pitch shifting, and more.
\n\nLink: http://ceur-ws.org/Vol-1609/16090547.pdf
\n\nThesis - Identifying Birds by Sound: Large-scale Acoustic Event Recognition for Avian Activity Monitoring by Stefan Kahl - Probably one of the most thorough resources on this page (A 200+ page thesis). Has an enormous amount of background information and an incredible amount of detail around different specific features and NN arhitectures as they deal with this specific problem.
\n\nLink: https://monarch.qucosa.de/api/qucosa%3A36986/attachment/ATT-0/
\n\nCornell's BirdNet - Cornell has already been reseraching this problem intensively and has deployed several ANN based classifiers online. It may be a good idea to download the app and play around with this to get a feel for what we are trying to do with longer, noisier, and more varied field recordings.\nhttps://birdnet.cornell.edu
\n\n-\nAgnieszka Mikołajczyk - has already put together an amazing set of resources on bird song recognition, many of which are reproduced here as well with more elaboration. - Link: https://github.com/AgaMiko/bird-recognition-review
\n\n-
\n\nAutomatic acoustic detection of birds through deep learning: the first Bird Audio Detection challenge - Used different datasets sourced from birds across Europe. Used Mel Frequency Spectral Coefficients (MFCCs) and Gaussian Mixture Models as a baseline method. Also used
\n\n-
\n\nA Baseline for Large-Scale Bird Species Identification in Field Recordings - Used the BirdCLEF challenge data and a CNN on spectrograms. Used MEL-scale log-amplitude spectrograms.. Trained on ResNet variations and had a goal of applying to mobile devices (and hence tried to use computationally inexpensive techniques).
\n\nLink: http://ceur-ws.org/Vol-2125/paper_85.pdf
\n\n-
\n\nBird Sound Recognition Using a Convolutional Neural Network - Also used Xeno Canto data through pretrained CNN models. Inputs were spectrograms. Limitation: seemed to work best with a low number of predicted classes (bird species).
\n\n\n\n-
\n\nBird Sound Classification using Convolutional Neural Networks - Applied similar process as above to Xeno Canto bird recordings using ResNet and Inception-v3 models.
\nLink: http://ceur-ws.org/Vol-2380/paper_68.pdf
Gabriel Preda also has some great resources here (some overlap) - https://www.kaggle.com/c/birdsong-recognition/discussion/160454
", "rawMarkdown": "Gabriel Preda also has some great resources here (some overlap) - https://www.kaggle.com/c/birdsong-recognition/discussion/160454", "votes": null }, { "id": "907200", "postDate": "06/29/2020 18:51:38", "content": "That's some great resources , thanks for the overview.
", "rawMarkdown": "That's some great resources , thanks for the overview.", "votes": null }, { "id": "907290", "postDate": "06/29/2020 20:18:45", "content": "Thanks Yash! Good luck to you!
", "rawMarkdown": "Thanks Yash! Good luck to you!", "votes": null }, { "id": "907435", "postDate": "06/30/2020 00:52:58", "content": "Also @louise2001 's fantastic review - https://www.kaggle.com/c/birdsong-recognition/discussion/158933
", "rawMarkdown": "Also @louise2001 's fantastic review - [https://www.kaggle.com/c/birdsong-recognition/discussion/158933](https://www.kaggle.com/c/birdsong-recognition/discussion/158933)", "votes": null }, { "id": "907578", "postDate": "06/30/2020 03:31:01", "content": "Checked the links! truly informative resources! thanks
", "rawMarkdown": "Checked the links! truly informative resources! thanks", "votes": null }, { "id": "973961", "postDate": "08/17/2020 16:23:25", "content": "\n\nAutomatic acoustic detection of birds through deep learning: the first Bird Audio Detection challenge - Used different datasets sourced from birds across Europe. Used Mel Frequency Spectral Coefficients (MFCCs) and Gaussian Mixture Models as a baseline method. Also used
\n-
\n
Missing link? https://arxiv.org/abs/1807.05812
", "rawMarkdown": "> Automatic acoustic detection of birds through deep learning: the first Bird Audio Detection challenge - Used different datasets sourced from birds across Europe. Used Mel Frequency Spectral Coefficients (MFCCs) and Gaussian Mixture Models as a baseline method. Also used\n>\n> -\n\nMissing link? https://arxiv.org/abs/1807.05812", "votes": null }, { "id": "1663337", "postDate": "01/25/2022 02:42:34", "content": "Thank you for the beautiful resources. Very informative and inspiring!
", "rawMarkdown": "Thank you for the beautiful resources. Very informative and inspiring!", "votes": null } ], "comments": [ { "id": 973961, "author_name": "marcogorelli", "author_url": "", "post_date": "08/17/2020 16:23:25", "content": "\n\nAutomatic acoustic detection of birds through deep learning: the first Bird Audio Detection challenge - Used different datasets sourced from birds across Europe. Used Mel Frequency Spectral Coefficients (MFCCs) and Gaussian Mixture Models as a baseline method. Also used
\n-
\n
Missing link? https://arxiv.org/abs/1807.05812
", "votes": null, "replies": [] }, { "id": 1663337, "author_name": "jananiharshu", "author_url": "", "post_date": "01/25/2022 02:42:34", "content": "Thank you for the beautiful resources. Very informative and inspiring!
", "votes": null, "replies": [] }, { "id": 907192, "author_name": "tpmeli", "author_url": "", "post_date": "06/29/2020 18:43:45", "content": "Gabriel Preda also has some great resources here (some overlap) - https://www.kaggle.com/c/birdsong-recognition/discussion/160454
", "votes": null, "replies": [] }, { "id": 907200, "author_name": "yash612", "author_url": "", "post_date": "06/29/2020 18:51:38", "content": "That's some great resources , thanks for the overview.
", "votes": null, "replies": [ { "id": 907290, "author_name": "tpmeli", "author_url": "", "post_date": "06/29/2020 20:18:45", "content": "Thanks Yash! Good luck to you!
", "votes": null, "replies": [] } ] }, { "id": 907435, "author_name": "tpmeli", "author_url": "", "post_date": "06/30/2020 00:52:58", "content": "Also @louise2001 's fantastic review - https://www.kaggle.com/c/birdsong-recognition/discussion/158933
", "votes": null, "replies": [] }, { "id": 907578, "author_name": "gauravdahiya", "author_url": "", "post_date": "06/30/2020 03:31:01", "content": "Checked the links! truly informative resources! thanks
", "votes": null, "replies": [] } ], "raw_markdown_by_id": { "907183": "# Literature Review #\nHere is a set of resources I found very useful. Feel free to add other resources you find in the comments and I will add to this resource and mention you. Thanks!\n\n-\n\n**Audio Based Bird Species Identification using Deep Learning Techniques** - This paper was one of the first to use deep learning methods over KNN and tree methods. It also elaborates on some really nice data preparation methods and is a fantastic introduction to what we will be doing here. It also elaborates on feature generation such as time shifting, pitch shifting, and more.\n\nLink: http://ceur-ws.org/Vol-1609/16090547.pdf\n\n**Thesis** - Identifying Birds by Sound: Large-scale Acoustic Event Recognition for Avian Activity Monitoring by Stefan Kahl - Probably one of the most thorough resources on this page (A 200+ page thesis). Has an enormous amount of background information and an incredible amount of detail around different specific features and NN arhitectures as they deal with this specific problem.\n\nLink: https://monarch.qucosa.de/api/qucosa%3A36986/attachment/ATT-0/\n\n**Cornell's BirdNet** - Cornell has already been reseraching this problem intensively and has deployed several ANN based classifiers online. It may be a good idea to download the app and play around with this to get a feel for what we are trying to do with longer, noisier, and more varied field recordings.\nhttps://birdnet.cornell.edu\n\n-\n**Agnieszka Mikołajczyk** - has already put together an amazing set of resources on bird song recognition, many of which are reproduced here as well with more elaboration. - Link: https://github.com/AgaMiko/bird-recognition-review\n\n-\n\n**Automatic acoustic detection of birds through deep learning: the first Bird Audio Detection challenge** - Used different datasets sourced from birds across Europe. Used Mel Frequency Spectral Coefficients (MFCCs) and Gaussian Mixture Models as a baseline method. Also used\n\n-\n\n**A Baseline for Large-Scale Bird Species Identification in Field Recordings** - Used the BirdCLEF challenge data and a CNN on spectrograms. Used [MEL-scale log-amplitude spectrograms.](https://towardsdatascience.com/getting-to-know-the-mel-spectrogram-31bca3e2d9d0). Trained on ResNet variations and had a goal of applying to mobile devices (and hence tried to use computationally inexpensive techniques).\n\nLink: http://ceur-ws.org/Vol-2125/paper_85.pdf\n\n-\n\n**Bird Sound Recognition Using a Convolutional Neural Network** - Also used Xeno Canto data through pretrained CNN models. Inputs were spectrograms. Limitation: seemed to work best with a low number of predicted classes (bird species). \n\nLink: https://www.researchgate.net/publication/328836649_Bird_Sound_Recognition_Using_a_Convolutional_Neural_Network\n\n\n-\n\n**Bird Sound Classification using Convolutional Neural Networks** - Applied similar process as above to Xeno Canto bird recordings using ResNet and Inception-v3 models. \nLink: http://ceur-ws.org/Vol-2380/paper_68.pdf", "907192": "Gabriel Preda also has some great resources here (some overlap) - https://www.kaggle.com/c/birdsong-recognition/discussion/160454", "907200": "That's some great resources , thanks for the overview.", "907290": "Thanks Yash! Good luck to you!", "907435": "Also @louise2001 's fantastic review - [https://www.kaggle.com/c/birdsong-recognition/discussion/158933](https://www.kaggle.com/c/birdsong-recognition/discussion/158933)", "907578": "Checked the links! truly informative resources! thanks", "973961": "> Automatic acoustic detection of birds through deep learning: the first Bird Audio Detection challenge - Used different datasets sourced from birds across Europe. Used Mel Frequency Spectral Coefficients (MFCCs) and Gaussian Mixture Models as a baseline method. Also used\n>\n> -\n\nMissing link? https://arxiv.org/abs/1807.05812", "1663337": "Thank you for the beautiful resources. Very informative and inspiring!" }, "source": "meta" }