How is everyone getting the audio data from the videos? Using google, everyone seems to use ffmpeg or moviepy, neither of which are installed packages. Also, I've seen a couple threads where people are able to install ffmpeg but I believe that goes against the rules for the competition.
", "messages": [ { "id": "702571", "postDate": "12/24/2019 21:10:29", "content": "How is everyone getting the audio data from the videos? Using google, everyone seems to use ffmpeg or moviepy, neither of which are installed packages. Also, I've seen a couple threads where people are able to install ffmpeg but I believe that goes against the rules for the competition.
", "rawMarkdown": "How is everyone getting the audio data from the videos? Using google, everyone seems to use ffmpeg or moviepy, neither of which are installed packages. Also, I've seen a couple threads where people are able to install ffmpeg but I believe that goes against the rules for the competition.", "votes": null }, { "id": "702594", "postDate": "12/24/2019 22:08:12", "content": "To add on, even though I am using tensorflow 2, I noticed that one of the installed packages is torchvision. However, torchvision cannot read the video files because of a different missing package, PyAV
", "rawMarkdown": "To add on, even though I am using tensorflow 2, I noticed that one of the installed packages is torchvision. However, torchvision cannot read the video files because of a different missing package, PyAV", "votes": null }, { "id": "702662", "postDate": "12/25/2019 01:28:29", "content": "It is not against the rules to install ffmpeg, or any other package, as long as you have the license to use (which in the case of ffmpeg is ok since its a free open source software). You just have to disclose in this thread and you can add the package as an external dataset to your kernel (within the 1gb limit) and install it.
", "rawMarkdown": "It is not against the rules to install ffmpeg, or any other package, as long as you have the license to use (which in the case of ffmpeg is ok since its a free open source software). You just have to disclose in [this thread](https://www.kaggle.com/c/deepfake-detection-challenge/discussion/121203) and you can add the package as an external dataset to your kernel (within the 1gb limit) and install it.", "votes": null }, { "id": "704040", "postDate": "12/27/2019 01:08:49", "content": "Yeah, it would be nice if PyAV is added to a default package in kaggle notebooks. https://www.kaggle.com/rakibilly/extract-audio-starter this notebook uses ffmpeg to convert video file to audio file, but it is not very customizable since the process is done outside of python. Even torchvision uses PyAV.
", "rawMarkdown": "Yeah, it would be nice if PyAV is added to a default package in kaggle notebooks. https://www.kaggle.com/rakibilly/extract-audio-starter this notebook uses ffmpeg to convert video file to audio file, but it is not very customizable since the process is done outside of python. Even torchvision uses PyAV.", "votes": null } ], "comments": [ { "id": 702594, "author_name": "wmol14", "author_url": "", "post_date": "12/24/2019 22:08:12", "content": "To add on, even though I am using tensorflow 2, I noticed that one of the installed packages is torchvision. However, torchvision cannot read the video files because of a different missing package, PyAV
", "votes": null, "replies": [ { "id": 704040, "author_name": "harangdev", "author_url": "", "post_date": "12/27/2019 01:08:49", "content": "Yeah, it would be nice if PyAV is added to a default package in kaggle notebooks. https://www.kaggle.com/rakibilly/extract-audio-starter this notebook uses ffmpeg to convert video file to audio file, but it is not very customizable since the process is done outside of python. Even torchvision uses PyAV.
", "votes": null, "replies": [] } ] }, { "id": 702662, "author_name": "pedromb", "author_url": "", "post_date": "12/25/2019 01:28:29", "content": "It is not against the rules to install ffmpeg, or any other package, as long as you have the license to use (which in the case of ffmpeg is ok since its a free open source software). You just have to disclose in this thread and you can add the package as an external dataset to your kernel (within the 1gb limit) and install it.
", "votes": null, "replies": [] } ], "raw_markdown_by_id": { "702571": "How is everyone getting the audio data from the videos? Using google, everyone seems to use ffmpeg or moviepy, neither of which are installed packages. Also, I've seen a couple threads where people are able to install ffmpeg but I believe that goes against the rules for the competition.", "702594": "To add on, even though I am using tensorflow 2, I noticed that one of the installed packages is torchvision. However, torchvision cannot read the video files because of a different missing package, PyAV", "702662": "It is not against the rules to install ffmpeg, or any other package, as long as you have the license to use (which in the case of ffmpeg is ok since its a free open source software). You just have to disclose in [this thread](https://www.kaggle.com/c/deepfake-detection-challenge/discussion/121203) and you can add the package as an external dataset to your kernel (within the 1gb limit) and install it.", "704040": "Yeah, it would be nice if PyAV is added to a default package in kaggle notebooks. https://www.kaggle.com/rakibilly/extract-audio-starter this notebook uses ffmpeg to convert video file to audio file, but it is not very customizable since the process is done outside of python. Even torchvision uses PyAV." }, "source": "meta" }