I just made public a large Mel Spectrogram dataset:
\nhttps://www.kaggle.com/datasets/richolson/birdclef-2024-big-bird-193k-mel-spectrograms
\nOver 193,000 224x224 Mel Spectrograms for your model-training-adventures.
\nI've also created a notebook that lets you preview Mel images and audio for all species:
\nhttps://www.kaggle.com/code/richolson/big-bird-spectrograms-look-and-listen
\n(if you only see samples for a few species - give it a minute to load - it's a big!)
\nI actually find browsing through the images and audio to be an enjoyable experience. A nice break from pretending to know anything about ML / data-science.
\nA much smaller version of this dataset is available here:
\nhttps://www.kaggle.com/datasets/richolson/birdclef-2024-mel-spectrograms
\n(if you are training your model for the first time - I'd honestly start with the small version)
Both of these were generated with:
\nhttps://www.kaggle.com/code/richolson/birdclef-2024-simple-mel-spectrogram-generator
Hope this is useful to someone!
\n-Rich
", "messages": [ { "id": "2745685", "postDate": "04/10/2024 18:41:12", "content": "I just made public a large Mel Spectrogram dataset:
\nhttps://www.kaggle.com/datasets/richolson/birdclef-2024-big-bird-193k-mel-spectrograms
\nOver 193,000 224x224 Mel Spectrograms for your model-training-adventures.
\nI've also created a notebook that lets you preview Mel images and audio for all species:
\nhttps://www.kaggle.com/code/richolson/big-bird-spectrograms-look-and-listen
\n(if you only see samples for a few species - give it a minute to load - it's a big!)
\nI actually find browsing through the images and audio to be an enjoyable experience. A nice break from pretending to know anything about ML / data-science.
\nA much smaller version of this dataset is available here:
\nhttps://www.kaggle.com/datasets/richolson/birdclef-2024-mel-spectrograms
\n(if you are training your model for the first time - I'd honestly start with the small version)
Both of these were generated with:
\nhttps://www.kaggle.com/code/richolson/birdclef-2024-simple-mel-spectrogram-generator
Hope this is useful to someone!
\n-Rich
", "rawMarkdown": "I just made public a large Mel Spectrogram dataset:\n\nhttps://www.kaggle.com/datasets/richolson/birdclef-2024-big-bird-193k-mel-spectrograms\n\nOver 193,000 224x224 Mel Spectrograms for your model-training-adventures.\n\nI've also created a notebook that lets you preview Mel images and audio for all species:\n\nhttps://www.kaggle.com/code/richolson/big-bird-spectrograms-look-and-listen\n\n(if you only see samples for a few species - give it a minute to load - it's a big!)\n\nI actually find browsing through the images and audio to be an enjoyable experience. A nice break from pretending to know anything about ML / data-science.\n\nA much smaller version of this dataset is available here:\nhttps://www.kaggle.com/datasets/richolson/birdclef-2024-mel-spectrograms\n(if you are training your model for the first time - I'd honestly start with the small version)\n\nBoth of these were generated with:\nhttps://www.kaggle.com/code/richolson/birdclef-2024-simple-mel-spectrogram-generator\n\nHope this is useful to someone!\n\n-Rich", "votes": null }, { "id": "2746926", "postDate": "04/11/2024 14:40:56", "content": "Could you please clarify the reason to choose 224x224 size? Thank you!
", "rawMarkdown": "Could you please clarify the reason to choose 224x224 size? Thank you!", "votes": null }, { "id": "2747143", "postDate": "04/11/2024 17:40:48", "content": "it's a common input size for small ImageNets:
\nhttps://keras.io/api/applications/mobilenet/
\nIf you'd like a different size -you can easily make your own dataset using:
\nhttps://www.kaggle.com/code/richolson/birdclef-2024-simple-mel-spectrogram-generator
Thank you @richolson
", "rawMarkdown": "Thank you @richolson", "votes": null } ], "comments": [ { "id": 2746926, "author_name": "samvelkoch", "author_url": "", "post_date": "04/11/2024 14:40:56", "content": "Could you please clarify the reason to choose 224x224 size? Thank you!
", "votes": null, "replies": [ { "id": 2747143, "author_name": "richolson", "author_url": "", "post_date": "04/11/2024 17:40:48", "content": "it's a common input size for small ImageNets:
\nhttps://keras.io/api/applications/mobilenet/
\nIf you'd like a different size -you can easily make your own dataset using:
\nhttps://www.kaggle.com/code/richolson/birdclef-2024-simple-mel-spectrogram-generator
Thank you @richolson
", "votes": null, "replies": [] } ] } ] } ], "raw_markdown_by_id": { "2745685": "I just made public a large Mel Spectrogram dataset:\n\nhttps://www.kaggle.com/datasets/richolson/birdclef-2024-big-bird-193k-mel-spectrograms\n\nOver 193,000 224x224 Mel Spectrograms for your model-training-adventures.\n\nI've also created a notebook that lets you preview Mel images and audio for all species:\n\nhttps://www.kaggle.com/code/richolson/big-bird-spectrograms-look-and-listen\n\n(if you only see samples for a few species - give it a minute to load - it's a big!)\n\nI actually find browsing through the images and audio to be an enjoyable experience. A nice break from pretending to know anything about ML / data-science.\n\nA much smaller version of this dataset is available here:\nhttps://www.kaggle.com/datasets/richolson/birdclef-2024-mel-spectrograms\n(if you are training your model for the first time - I'd honestly start with the small version)\n\nBoth of these were generated with:\nhttps://www.kaggle.com/code/richolson/birdclef-2024-simple-mel-spectrogram-generator\n\nHope this is useful to someone!\n\n-Rich", "2746926": "Could you please clarify the reason to choose 224x224 size? Thank you!", "2747143": "it's a common input size for small ImageNets:\n\nhttps://keras.io/api/applications/mobilenet/\n\nIf you'd like a different size -you can easily make your own dataset using: \nhttps://www.kaggle.com/code/richolson/birdclef-2024-simple-mel-spectrogram-generator", "2747184": "Thank you @richolson" }, "source": "meta" }