I submitted my audio model yesterday,but I got 0.69314 in the end. At first, I thought it was beacause I only used audio model, and the number of videos with fake audio is much smaller than number of videos with fake pictures(fake faces). But today , when I commit my notebook, I found it took me 7mins for 400 files and it only took few min for submitting. So I wondered if there was something wrong when i processed the file. So do you have any advice for audio-file processing? I use ffmpeg to separate the audio from video.
", "messages": [ { "id": 745605, "postDate": "2020-02-14T01:36:46.787Z", "content": "I submitted my audio model yesterday,but I got 0.69314 in the end. At first, I thought it was beacause I only used audio model, and the number of videos with fake audio is much smaller than number of videos with fake pictures(fake faces). But today , when I commit my notebook, I found it took me 7mins for 400 files and it only took few min for submitting. So I wondered if there was something wrong when i processed the file. So do you have any advice for audio-file processing? I use ffmpeg to separate the audio from video.
", "rawMarkdown": "I submitted my audio model yesterday,but I got 0.69314 in the end. At first, I thought it was beacause I only used audio model, and the number of videos with fake audio is much smaller than number of videos with fake pictures(fake faces). But today , when I commit my notebook, I found it took me 7mins for 400 files and it only took few min for submitting. So I wondered if there was something wrong when i processed the file. So do you have any advice for audio-file processing? I use ffmpeg to separate the audio from video.", "votes": 1 }, { "id": 745720, "postDate": "2020-02-14T05:47:39.810Z", "content": "You can do the same old os.walk for your output folder and print results in the notebook or save is as a csv or something.\nYou can also compare the initial list of files and your audios, whether their lengths match or not.
", "rawMarkdown": "You can do the same old os.walk for your output folder and print results in the notebook or save is as a csv or something.\nYou can also compare the initial list of files and your audios, whether their lengths match or not.", "replies": [ { "id": 745759, "postDate": "2020-02-14T06:57:27.290Z", "content": "thx,defileroff. I did find audio files in \"working\" folder. And I wrote the path like\"../working/xxxx\". But something still was wrong with my code since I got 0.69314 again,which indicated that I triggered my exception code.
", "rawMarkdown": "thx,defileroff. I did find audio files in \"working\" folder. And I wrote the path like\"../working/xxxx\". But something still was wrong with my code since I got 0.69314 again,which indicated that I triggered my exception code." }, { "id": 749309, "postDate": "2020-02-18T15:07:15.817Z", "content": "You can set a fixed prediction score for files with errors. Make sure that your code runs well after you've committed it. By the way, there are some corrupt videos in the full test set, maybe that's causing a problem.
", "rawMarkdown": "You can set a fixed prediction score for files with errors. Make sure that your code runs well after you've committed it. By the way, there are some corrupt videos in the full test set, maybe that's causing a problem." } ] }, { "id": 745610, "postDate": "2020-02-14T01:51:25.410Z", "content": "so how to read files in output folder?
", "rawMarkdown": "so how to read files in output folder?" }, { "id": 749003, "postDate": "2020-02-18T06:50:33.480Z", "rawMarkdown": "", "isDeleted": true } ], "comments": [ { "id": 745720, "author_name": "petya", "author_url": "", "post_date": "2020-02-14T05:47:39.810000", "content": "You can do the same old os.walk for your output folder and print results in the notebook or save is as a csv or something.\nYou can also compare the initial list of files and your audios, whether their lengths match or not.
", "votes": 0, "replies": [ { "id": 745759, "author_name": "Emmettj", "author_url": "", "post_date": "2020-02-14T06:57:27.290000", "content": "thx,defileroff. I did find audio files in \"working\" folder. And I wrote the path like\"../working/xxxx\". But something still was wrong with my code since I got 0.69314 again,which indicated that I triggered my exception code.
", "votes": 0, "replies": [] }, { "id": 749309, "author_name": "petya", "author_url": "", "post_date": "2020-02-18T15:07:15.817000", "content": "You can set a fixed prediction score for files with errors. Make sure that your code runs well after you've committed it. By the way, there are some corrupt videos in the full test set, maybe that's causing a problem.
", "votes": 0, "replies": [] } ] }, { "id": 745610, "author_name": "Emmettj", "author_url": "", "post_date": "2020-02-14T01:51:25.410000", "content": "so how to read files in output folder?
", "votes": 0, "replies": [] }, { "id": 749003, "author_name": "", "author_url": "", "post_date": "2020-02-18T06:50:33.480000", "content": "", "votes": 0, "replies": [] } ], "raw_markdown_by_id": { "745605": "I submitted my audio model yesterday,but I got 0.69314 in the end. At first, I thought it was beacause I only used audio model, and the number of videos with fake audio is much smaller than number of videos with fake pictures(fake faces). But today , when I commit my notebook, I found it took me 7mins for 400 files and it only took few min for submitting. So I wondered if there was something wrong when i processed the file. So do you have any advice for audio-file processing? I use ffmpeg to separate the audio from video.", "745720": "You can do the same old os.walk for your output folder and print results in the notebook or save is as a csv or something.\nYou can also compare the initial list of files and your audios, whether their lengths match or not.", "745610": "so how to read files in output folder?", "749003": "" } }