Now that this competition is over, many of us might be asking ourselves, what's next? I found myself being more interested in audio classification projects. I did a websearch and found some links to really interesting audio datasets. Check them out:
\nXeno-Canto: The famous Xeno-canto recordings. Thinking of scraping for a few classes, and refining the methods used in this competition to improve accuracy of birdcall detection.
\nDigit Sounds Recordings: The MNIST of audio. I uploaded a version on Kaggle: https://www.kaggle.com/alanchn31/free-spoken-digits
\nAudioSet: an expanding ontology of 632 audio event classes and a collection of 2,084,320 human-labeled 10-second sound clips drawn from YouTube videos
\nMillion Song Dataset: This is a freely-available collection of audio features and metadata for a million contemporary popular music tracks.
\nDCASE Detection and Classification of
\nAcoustic Scenes and Events, computational analysis of sound events and scene analysis (Thanks @hidehisaarai1213)
Speech Commands Dataset 65,000 one-second long utterances of 30 short words, by thousands of different people (Thanks @nyleve)
\nI'll update if I find more interesting datasets. Do provide some recommendations as well if you have any. Happy Kaggling everyone!
", "messages": [ { "id": "1015145", "postDate": "09/18/2020 01:50:01", "content": "Now that this competition is over, many of us might be asking ourselves, what's next? I found myself being more interested in audio classification projects. I did a websearch and found some links to really interesting audio datasets. Check them out:
\nXeno-Canto: The famous Xeno-canto recordings. Thinking of scraping for a few classes, and refining the methods used in this competition to improve accuracy of birdcall detection.
\nDigit Sounds Recordings: The MNIST of audio. I uploaded a version on Kaggle: https://www.kaggle.com/alanchn31/free-spoken-digits
\nAudioSet: an expanding ontology of 632 audio event classes and a collection of 2,084,320 human-labeled 10-second sound clips drawn from YouTube videos
\nMillion Song Dataset: This is a freely-available collection of audio features and metadata for a million contemporary popular music tracks.
\nDCASE Detection and Classification of
\nAcoustic Scenes and Events, computational analysis of sound events and scene analysis (Thanks @hidehisaarai1213)
Speech Commands Dataset 65,000 one-second long utterances of 30 short words, by thousands of different people (Thanks @nyleve)
\nI'll update if I find more interesting datasets. Do provide some recommendations as well if you have any. Happy Kaggling everyone!
", "rawMarkdown": "Now that this competition is over, many of us might be asking ourselves, what's next? I found myself being more interested in audio classification projects. I did a websearch and found some links to really interesting audio datasets. Check them out:\n\n[Xeno-Canto](https://www.xeno-canto.org/): The famous Xeno-canto recordings. Thinking of scraping for a few classes, and refining the methods used in this competition to improve accuracy of birdcall detection.\n\n[Digit Sounds Recordings](https://github.com/Jakobovski/free-spoken-digit-dataset): The MNIST of audio. I uploaded a version on Kaggle: https://www.kaggle.com/alanchn31/free-spoken-digits\n\n[AudioSet](https://research.google.com/audioset/): an expanding ontology of 632 audio event classes and a collection of 2,084,320 human-labeled 10-second sound clips drawn from YouTube videos\n\n[Million Song Dataset](https://labrosa.ee.columbia.edu/millionsong/): This is a freely-available collection of audio features and metadata for a million contemporary popular music tracks.\n\n[DCASE](http://dcase.community/) Detection and Classification of\nAcoustic Scenes and Events, computational analysis of sound events and scene analysis (Thanks @hidehisaarai1213) \n\n[Speech Commands Dataset] (https://ai.googleblog.com/2017/08/launching-speech-commands-dataset.html) 65,000 one-second long utterances of 30 short words, by thousands of different people (Thanks @nyleve)\n\nI'll update if I find more interesting datasets. Do provide some recommendations as well if you have any. Happy Kaggling everyone!", "votes": null }, { "id": "1015165", "postDate": "09/18/2020 02:32:40", "content": "Thank you so much for uploading these links. I have always been intrigued by audio classification and all the papers written on it. Have a nice day!
", "rawMarkdown": "Thank you so much for uploading these links. I have always been intrigued by audio classification and all the papers written on it. Have a nice day!", "votes": null }, { "id": "1015189", "postDate": "09/18/2020 03:06:58", "content": "No problem! Audio is a really interesting field with potential to develop quickly in the short term. I am trying to get more practice in this area as well. Thanks for your interest.
", "rawMarkdown": "No problem! Audio is a really interesting field with potential to develop quickly in the short term. I am trying to get more practice in this area as well. Thanks for your interest.", "votes": null }, { "id": "1015352", "postDate": "09/18/2020 06:16:56", "content": "Google has a nice speech command dataset here :
\nhttps://ai.googleblog.com/2017/08/launching-speech-commands-dataset.html
\n--> 65,000 one-second long utterances of 30 short words, by thousands of different people
\nIt was partially used in kaggle competition 3 years back :
\nhttps://www.kaggle.com/c/tensorflow-speech-recognition-challenge
Every year, DCASE community holds a number of competitions. Personally I have never participated in those so far, but been interested in them after Freesound Audio Tagging 2019 last year (which was the first audio competition for me.)
", "rawMarkdown": "Every year, [DCASE](http://dcase.community/) community holds a number of competitions. Personally I have never participated in those so far, but been interested in them after Freesound Audio Tagging 2019 last year (which was the first audio competition for me.)", "votes": null }, { "id": "1016028", "postDate": "09/18/2020 15:55:33", "content": "Thank you! Added it to the list. DCASE looks interesting, there are multiple challenges available. Definitely worth a look!
", "rawMarkdown": "Thank you! Added it to the list. DCASE looks interesting, there are multiple challenges available. Definitely worth a look!", "votes": null }, { "id": "1016031", "postDate": "09/18/2020 15:59:12", "content": "Thank you, this looks interesting! I have added it to the list. I tried implementing some solutions used in tensorflow speech recognition challenge for this competition as well, such as focalloss and f1loss. Speech detection has a nice tie-in with NLP which makes it really interesting.
", "rawMarkdown": "Thank you, this looks interesting! I have added it to the list. I tried implementing some solutions used in tensorflow speech recognition challenge for this competition as well, such as focalloss and f1loss. Speech detection has a nice tie-in with NLP which makes it really interesting.", "votes": null } ], "comments": [ { "id": 1015165, "author_name": "rishikeshkanabar", "author_url": "", "post_date": "09/18/2020 02:32:40", "content": "Thank you so much for uploading these links. I have always been intrigued by audio classification and all the papers written on it. Have a nice day!
", "votes": null, "replies": [ { "id": 1015189, "author_name": "alanchn31", "author_url": "", "post_date": "09/18/2020 03:06:58", "content": "No problem! Audio is a really interesting field with potential to develop quickly in the short term. I am trying to get more practice in this area as well. Thanks for your interest.
", "votes": null, "replies": [] } ] }, { "id": 1015352, "author_name": "nyleve", "author_url": "", "post_date": "09/18/2020 06:16:56", "content": "Google has a nice speech command dataset here :
\nhttps://ai.googleblog.com/2017/08/launching-speech-commands-dataset.html
\n--> 65,000 one-second long utterances of 30 short words, by thousands of different people
\nIt was partially used in kaggle competition 3 years back :
\nhttps://www.kaggle.com/c/tensorflow-speech-recognition-challenge
Thank you, this looks interesting! I have added it to the list. I tried implementing some solutions used in tensorflow speech recognition challenge for this competition as well, such as focalloss and f1loss. Speech detection has a nice tie-in with NLP which makes it really interesting.
", "votes": null, "replies": [] } ] }, { "id": 1015772, "author_name": "hidehisaarai1213", "author_url": "", "post_date": "09/18/2020 12:07:49", "content": "Every year, DCASE community holds a number of competitions. Personally I have never participated in those so far, but been interested in them after Freesound Audio Tagging 2019 last year (which was the first audio competition for me.)
", "votes": null, "replies": [ { "id": 1016028, "author_name": "alanchn31", "author_url": "", "post_date": "09/18/2020 15:55:33", "content": "Thank you! Added it to the list. DCASE looks interesting, there are multiple challenges available. Definitely worth a look!
", "votes": null, "replies": [] } ] } ], "raw_markdown_by_id": { "1015145": "Now that this competition is over, many of us might be asking ourselves, what's next? I found myself being more interested in audio classification projects. I did a websearch and found some links to really interesting audio datasets. Check them out:\n\n[Xeno-Canto](https://www.xeno-canto.org/): The famous Xeno-canto recordings. Thinking of scraping for a few classes, and refining the methods used in this competition to improve accuracy of birdcall detection.\n\n[Digit Sounds Recordings](https://github.com/Jakobovski/free-spoken-digit-dataset): The MNIST of audio. I uploaded a version on Kaggle: https://www.kaggle.com/alanchn31/free-spoken-digits\n\n[AudioSet](https://research.google.com/audioset/): an expanding ontology of 632 audio event classes and a collection of 2,084,320 human-labeled 10-second sound clips drawn from YouTube videos\n\n[Million Song Dataset](https://labrosa.ee.columbia.edu/millionsong/): This is a freely-available collection of audio features and metadata for a million contemporary popular music tracks.\n\n[DCASE](http://dcase.community/) Detection and Classification of\nAcoustic Scenes and Events, computational analysis of sound events and scene analysis (Thanks @hidehisaarai1213) \n\n[Speech Commands Dataset] (https://ai.googleblog.com/2017/08/launching-speech-commands-dataset.html) 65,000 one-second long utterances of 30 short words, by thousands of different people (Thanks @nyleve)\n\nI'll update if I find more interesting datasets. Do provide some recommendations as well if you have any. Happy Kaggling everyone!", "1015165": "Thank you so much for uploading these links. I have always been intrigued by audio classification and all the papers written on it. Have a nice day!", "1015189": "No problem! Audio is a really interesting field with potential to develop quickly in the short term. I am trying to get more practice in this area as well. Thanks for your interest.", "1015352": "Google has a nice speech command dataset here : \nhttps://ai.googleblog.com/2017/08/launching-speech-commands-dataset.html\n--> 65,000 one-second long utterances of 30 short words, by thousands of different people\nIt was partially used in kaggle competition 3 years back : \nhttps://www.kaggle.com/c/tensorflow-speech-recognition-challenge", "1015772": "Every year, [DCASE](http://dcase.community/) community holds a number of competitions. Personally I have never participated in those so far, but been interested in them after Freesound Audio Tagging 2019 last year (which was the first audio competition for me.)", "1016028": "Thank you! Added it to the list. DCASE looks interesting, there are multiple challenges available. Definitely worth a look!", "1016031": "Thank you, this looks interesting! I have added it to the list. I tried implementing some solutions used in tensorflow speech recognition challenge for this competition as well, such as focalloss and f1loss. Speech detection has a nice tie-in with NLP which makes it really interesting." }, "source": "meta" }