As a beginner of this competition, I did some review of the more state-of-the-art technic in recent months. Everythings from internet, its my first time writing review in Discussion, if there's any problem plz comment here, and I'll delete the wrong part.
\nSound Event Detection will be writen as SED in this game.
PANNs -- a popular algorithm in notebooks, which is named as Pretrained Audio Neural Networks. which is figure that CNN14 or Wavegram-Logmel-CNN get a well performance.
\n-- Ref: PANNs: Large-Scale Pretrained Audio Neural Networks for Audio Pattern Recognition
\n-- code & paper
One of similar competition FSDKaggle2018 by using CNN.
\n-- Ref: LEARNING SOUND EVENT CLASSIFIERS FROM WEB AUDIO WITH NOISY LABELS
\n-- code & paper
To improve your score could be used.
\n-- Ref: Sound Event Detection with Depthwise Separable and Dilated Convolutions
\n-- code & paper
Another world greatest SED competition. (DCASE 2019 Task 4) by using CNN.
\n-- Ref: Guided Learning Convolution System for DCASE 2019 Task 4
\n-- code & paper
As a beginner of this competition, I did some review of the more state-of-the-art technic in recent months. Everythings from internet, its my first time writing review in Discussion, if there's any problem plz comment here, and I'll delete the wrong part.
\nSound Event Detection will be writen as SED in this game.
PANNs -- a popular algorithm in notebooks, which is named as Pretrained Audio Neural Networks. which is figure that CNN14 or Wavegram-Logmel-CNN get a well performance.
\n-- Ref: PANNs: Large-Scale Pretrained Audio Neural Networks for Audio Pattern Recognition
\n-- code & paper
One of similar competition FSDKaggle2018 by using CNN.
\n-- Ref: LEARNING SOUND EVENT CLASSIFIERS FROM WEB AUDIO WITH NOISY LABELS
\n-- code & paper
To improve your score could be used.
\n-- Ref: Sound Event Detection with Depthwise Separable and Dilated Convolutions
\n-- code & paper
Another world greatest SED competition. (DCASE 2019 Task 4) by using CNN.
\n-- Ref: Guided Learning Convolution System for DCASE 2019 Task 4
\n-- code & paper
Good job! And we have a great implementation of SED inside this competition: https://www.kaggle.com/hidehisaarai1213/introduction-to-sound-event-detection
", "rawMarkdown": "Good job! And we have a great implementation of SED inside this competition: https://www.kaggle.com/hidehisaarai1213/introduction-to-sound-event-detection", "votes": null }, { "id": "972808", "postDate": "08/16/2020 20:50:12", "content": "I'll keep adding the paper here.
\n6.One of Golden place suggest using attention-based pooling at Notebook. I guess it could be a way to improve LB? I made a copy here.
\n -- Ref: Attention-based Deep Multiple Instance Learning -Maximilian Ilse
\n -- code & paper
\nA based study paper for understanding: Studying The Effect of MIL Pooling Filters on MIL Tasks
Ok, thanks! Maybe, it will be more convenient for you add or duplicate sources inside the topic, not only in comment.
", "rawMarkdown": "Ok, thanks! Maybe, it will be more convenient for you add or duplicate sources inside the topic, not only in comment.", "votes": null }, { "id": "977897", "postDate": "08/19/2020 19:32:42", "content": "7.A paper: The impact of environmental factors in birdsong acquisition using automated recorders
", "rawMarkdown": "7.A paper: [The impact of environmental factors in birdsong acquisition using automated recorders](https://onlinelibrary.wiley.com/doi/full/10.1002/ece3.3889)", "votes": null } ], "comments": [ { "id": 972754, "author_name": "koza4ukdmitrij", "author_url": "", "post_date": "08/16/2020 19:53:22", "content": "Good job! And we have a great implementation of SED inside this competition: https://www.kaggle.com/hidehisaarai1213/introduction-to-sound-event-detection
", "votes": null, "replies": [] }, { "id": 972808, "author_name": "dwchen", "author_url": "", "post_date": "08/16/2020 20:50:12", "content": "I'll keep adding the paper here.
\n6.One of Golden place suggest using attention-based pooling at Notebook. I guess it could be a way to improve LB? I made a copy here.
\n -- Ref: Attention-based Deep Multiple Instance Learning -Maximilian Ilse
\n -- code & paper
\nA based study paper for understanding: Studying The Effect of MIL Pooling Filters on MIL Tasks
Ok, thanks! Maybe, it will be more convenient for you add or duplicate sources inside the topic, not only in comment.
", "votes": null, "replies": [] } ] }, { "id": 977897, "author_name": "dwchen", "author_url": "", "post_date": "08/19/2020 19:32:42", "content": "7.A paper: The impact of environmental factors in birdsong acquisition using automated recorders
", "votes": null, "replies": [] } ], "raw_markdown_by_id": { "972740": "As a beginner of this competition, I did some review of the more state-of-the-art technic in recent months. Everythings from internet, its my first time writing review in Discussion, if there's any problem plz comment here, and I'll delete the wrong part.\n\n1. Sound Event Detection will be writen as SED in this game.\n\n2. PANNs -- a popular algorithm in notebooks, which is named as Pretrained Audio Neural Networks. which is figure that CNN14 or Wavegram-Logmel-CNN get a well performance.\n -- Ref: PANNs: Large-Scale Pretrained Audio Neural Networks for Audio Pattern Recognition\n -- [code](https://github.com/qiuqiangkong/audioset_tagging_cnn/) & [paper](https://arxiv.org/abs/1912.10211)\n\n\n3. One of similar competition [FSDKaggle2018](https://www.kaggle.com/c/freesound-audio-tagging) by using CNN.\n -- Ref: LEARNING SOUND EVENT CLASSIFIERS FROM WEB AUDIO WITH NOISY LABELS\n -- [code](https://github.com/lRomul/argus-freesound) & [paper](https://arxiv.org/abs/1901.01189v2)\n\n4. To improve your score could be used.\n -- Ref: Sound Event Detection with Depthwise Separable and Dilated Convolutions\n -- [code](https://github.com/dr-costas/dnd-sed) & [paper](https://arxiv.org/abs/2002.00476v1)\n\n5. Another world greatest SED competition. (DCASE 2019 Task 4) by using CNN.\n -- Ref: Guided Learning Convolution System for DCASE 2019 Task 4\n -- [code](https://github.com/Kikyo-16/Sound_event_detection) & [paper](https://arxiv.org/abs/1909.06178v1)", "972754": "Good job! And we have a great implementation of SED inside this competition: https://www.kaggle.com/hidehisaarai1213/introduction-to-sound-event-detection", "972808": "I'll keep adding the paper here.\n\n6.One of Golden place suggest using [attention-based pooling](https://www.kaggle.com/hidehisaarai1213/introduction-to-sound-event-detection?scriptVersionId=40371251) at Notebook. I guess it could be a way to improve LB? I made a copy here.\n -- Ref: Attention-based Deep Multiple Instance Learning -Maximilian Ilse\n -- [code](https://github.com/utayao/Atten_Deep_MIL/blob/master/utl/custom_layers.py) & [paper](https://arxiv.org/pdf/1802.04712.pdf)\nA based study paper for understanding: [Studying The Effect of MIL Pooling Filters on MIL Tasks](https://arxiv.org/abs/2006.01561)", "972815": "Ok, thanks! Maybe, it will be more convenient for you add or duplicate sources inside the topic, not only in comment.", "977897": "7.A paper: [The impact of environmental factors in birdsong acquisition using automated recorders](https://onlinelibrary.wiley.com/doi/full/10.1002/ece3.3889)" }, "source": "meta" }