Hello everyone,\n This is my first competition related to audio data. I am trying to create spectograms but the amount of time it is taking is several hours. I am wondering if it is normal behavior? Can we run cv2 and normal python code on GPU? I tried numba but it doesn't support many libraries. I hope to get some guidance on the same.
\n\nThank You for your time.
", "messages": [ { "id": "949382", "postDate": "07/28/2020 15:51:46", "content": "Hello everyone,\n This is my first competition related to audio data. I am trying to create spectograms but the amount of time it is taking is several hours. I am wondering if it is normal behavior? Can we run cv2 and normal python code on GPU? I tried numba but it doesn't support many libraries. I hope to get some guidance on the same.
\n\nThank You for your time.
", "rawMarkdown": "Hello everyone,\n This is my first competition related to audio data. I am trying to create spectograms but the amount of time it is taking is several hours. I am wondering if it is normal behavior? Can we run cv2 and normal python code on GPU? I tried numba but it doesn't support many libraries. I hope to get some guidance on the same.\n\nThank You for your time.", "votes": null }, { "id": "949465", "postDate": "07/28/2020 16:46:15", "content": "Hello Yash,\nyou could use precomputed spectrograms -> https://www.kaggle.com/ryches/birdsongspectrograms
\n\n... or you could use the GPU for spectrogram generation on the fly -> https://towardsdatascience.com/how-to-easily-process-audio-on-your-gpu-with-tensorflow-2d9d91360f06
", "rawMarkdown": "Hello Yash,\nyou could use precomputed spectrograms -> [https://www.kaggle.com/ryches/birdsongspectrograms](https://www.kaggle.com/ryches/birdsongspectrograms)\n\n... or you could use the GPU for spectrogram generation on the fly -> [https://towardsdatascience.com/how-to-easily-process-audio-on-your-gpu-with-tensorflow-2d9d91360f06](https://towardsdatascience.com/how-to-easily-process-audio-on-your-gpu-with-tensorflow-2d9d91360f06)", "votes": null }, { "id": "949646", "postDate": "07/28/2020 19:09:19", "content": "If you need the spectrogram for the whole audio the AgentAuers link is good, otherwise I have created spectrograms for every 5 second of the audio file resampled each a 32 kHz: notebook parallelized-spectrogram-as-numpy.\nAt this point if you don't want to do it to learn how to create spectrograms there are a already a lot of kernels and datasets with different specifications so you can use them directly.
", "rawMarkdown": "If you need the spectrogram for the whole audio the AgentAuers link is good, otherwise I have created spectrograms for every 5 second of the audio file resampled each a 32 kHz: [notebook parallelized-spectrogram-as-numpy](https://www.kaggle.com/pranavkasela/parallelized-spectrogram-as-numpy).\nAt this point if you don't want to do it to learn how to create spectrograms there are a already a lot of kernels and datasets with different specifications so you can use them directly.", "votes": null } ], "comments": [ { "id": 949465, "author_name": "agentauers", "author_url": "", "post_date": "07/28/2020 16:46:15", "content": "Hello Yash,\nyou could use precomputed spectrograms -> https://www.kaggle.com/ryches/birdsongspectrograms
\n\n... or you could use the GPU for spectrogram generation on the fly -> https://towardsdatascience.com/how-to-easily-process-audio-on-your-gpu-with-tensorflow-2d9d91360f06
", "votes": null, "replies": [] }, { "id": 949646, "author_name": "pranavkasela", "author_url": "", "post_date": "07/28/2020 19:09:19", "content": "If you need the spectrogram for the whole audio the AgentAuers link is good, otherwise I have created spectrograms for every 5 second of the audio file resampled each a 32 kHz: notebook parallelized-spectrogram-as-numpy.\nAt this point if you don't want to do it to learn how to create spectrograms there are a already a lot of kernels and datasets with different specifications so you can use them directly.
", "votes": null, "replies": [] } ], "raw_markdown_by_id": { "949382": "Hello everyone,\n This is my first competition related to audio data. I am trying to create spectograms but the amount of time it is taking is several hours. I am wondering if it is normal behavior? Can we run cv2 and normal python code on GPU? I tried numba but it doesn't support many libraries. I hope to get some guidance on the same.\n\nThank You for your time.", "949465": "Hello Yash,\nyou could use precomputed spectrograms -> [https://www.kaggle.com/ryches/birdsongspectrograms](https://www.kaggle.com/ryches/birdsongspectrograms)\n\n... or you could use the GPU for spectrogram generation on the fly -> [https://towardsdatascience.com/how-to-easily-process-audio-on-your-gpu-with-tensorflow-2d9d91360f06](https://towardsdatascience.com/how-to-easily-process-audio-on-your-gpu-with-tensorflow-2d9d91360f06)", "949646": "If you need the spectrogram for the whole audio the AgentAuers link is good, otherwise I have created spectrograms for every 5 second of the audio file resampled each a 32 kHz: [notebook parallelized-spectrogram-as-numpy](https://www.kaggle.com/pranavkasela/parallelized-spectrogram-as-numpy).\nAt this point if you don't want to do it to learn how to create spectrograms there are a already a lot of kernels and datasets with different specifications so you can use them directly." }, "source": "meta" }