{ "id": 127336, "title": "SAVI looks for inconsistencies in video to detect fakes", "url": "/competitions/deepfake-detection-challenge/discussion/127336", "author_name": "cyberia", "post_date": "2020-01-23T11:20:37.381000", "votes": 11, "comment_count": 6, "views": 0, "content": "

During literature review on Deepfakes I came across SAVI by SRI International.

\n\n
\n

SRI’s Spotting Audio-Visual Inconsistencies (SAVI) techniques detect tampered videos by identifying discrepancies between the audio and visual tracks. For example, the system can detect when lip synchronization is a little off or if there is an unexplained visual “jerk” in the video. Or it can flag a video as possibly tampered if the visual scene is outdoors, but analysis of the reverberation properties of the audio track indicates the recording was done in a small room.

\n
\n\n

See it in action https://www.youtube.com/watch?v=tm8PdqRGUYQ&feature=youtu.be

\n\n

\"\"

", "messages": [ { "id": 727007, "postDate": "2020-01-23T11:20:37.380Z", "content": "

During literature review on Deepfakes I came across SAVI by SRI International.

\n\n
\n

SRI’s Spotting Audio-Visual Inconsistencies (SAVI) techniques detect tampered videos by identifying discrepancies between the audio and visual tracks. For example, the system can detect when lip synchronization is a little off or if there is an unexplained visual “jerk” in the video. Or it can flag a video as possibly tampered if the visual scene is outdoors, but analysis of the reverberation properties of the audio track indicates the recording was done in a small room.

\n
\n\n

See it in action https://www.youtube.com/watch?v=tm8PdqRGUYQ&feature=youtu.be

\n\n

\"\"

", "rawMarkdown": "During literature review on Deepfakes I came across SAVI by SRI International.\n\n>SRI’s Spotting Audio-Visual Inconsistencies (SAVI) techniques detect tampered videos by identifying discrepancies between the audio and visual tracks. For example, the system can detect when lip synchronization is a little off or if there is an unexplained visual “jerk” in the video. Or it can flag a video as possibly tampered if the visual scene is outdoors, but analysis of the reverberation properties of the audio track indicates the recording was done in a small room.\n\nSee it in action https://www.youtube.com/watch?v=tm8PdqRGUYQ&feature=youtu.be\n\n![](https://www.googleapis.com/download/storage/v1/b/kaggle-user-content/o/inbox%2F551520%2Fb2bcd1d20c908ff1a6b1539f568f02a6%2FScreen%20Shot%202020-01-23%20at%2011.18.13%20AM.png?generation=1579778355998979&alt=media)\n", "votes": 10 }, { "id": 728715, "postDate": "2020-01-25T06:12:47.440Z", "content": "

Can you link us to the paper where this came from?

", "rawMarkdown": "Can you link us to the paper where this came from?" }, { "id": 729279, "postDate": "2020-01-26T00:55:50.783Z", "content": "

paper link: https://staff.fnwi.uva.nl/t.e.j.mensink/publications/bolles17cvprwmf.pdf

", "rawMarkdown": "paper link: https://staff.fnwi.uva.nl/t.e.j.mensink/publications/bolles17cvprwmf.pdf", "votes": 4, "isDeleted": true, "replies": [ { "id": 729376, "postDate": "2020-01-26T05:25:16.413Z", "content": "

Nice, thanks for this!

", "rawMarkdown": "Nice, thanks for this! " }, { "id": 729414, "postDate": "2020-01-26T07:01:35.333Z", "content": "

upvote my answer would be appreciated.

", "rawMarkdown": "upvote my answer would be appreciated. ", "votes": 1, "isDeleted": true }, { "id": 729419, "postDate": "2020-01-26T07:19:04.300Z", "content": "

Would a second make up for that?

", "rawMarkdown": "Would a second make up for that?" }, { "id": 729481, "postDate": "2020-01-26T09:01:15.343Z", "content": "

haha, thanks

", "rawMarkdown": "haha, thanks", "votes": 1, "isDeleted": true } ] } ], "comments": [ { "id": 728715, "author_name": "Untitled Algotithm", "author_url": "", "post_date": "2020-01-25T06:12:47.440000", "content": "

Can you link us to the paper where this came from?

", "votes": 0, "replies": [] }, { "id": 729279, "author_name": "", "author_url": "", "post_date": "2020-01-26T00:55:50.783000", "content": "

paper link: https://staff.fnwi.uva.nl/t.e.j.mensink/publications/bolles17cvprwmf.pdf

", "votes": 4, "replies": [ { "id": 729376, "author_name": "Untitled Algotithm", "author_url": "", "post_date": "2020-01-26T05:25:16.413000", "content": "

Nice, thanks for this!

", "votes": 0, "replies": [] }, { "id": 729414, "author_name": "", "author_url": "", "post_date": "2020-01-26T07:01:35.333000", "content": "

upvote my answer would be appreciated.

", "votes": 1, "replies": [] }, { "id": 729419, "author_name": "Untitled Algotithm", "author_url": "", "post_date": "2020-01-26T07:19:04.300000", "content": "

Would a second make up for that?

", "votes": 0, "replies": [] }, { "id": 729481, "author_name": "", "author_url": "", "post_date": "2020-01-26T09:01:15.343000", "content": "

haha, thanks

", "votes": 1, "replies": [] } ] } ], "raw_markdown_by_id": { "727007": "During literature review on Deepfakes I came across SAVI by SRI International.\n\n>SRI’s Spotting Audio-Visual Inconsistencies (SAVI) techniques detect tampered videos by identifying discrepancies between the audio and visual tracks. For example, the system can detect when lip synchronization is a little off or if there is an unexplained visual “jerk” in the video. Or it can flag a video as possibly tampered if the visual scene is outdoors, but analysis of the reverberation properties of the audio track indicates the recording was done in a small room.\n\nSee it in action https://www.youtube.com/watch?v=tm8PdqRGUYQ&feature=youtu.be\n\n![](https://www.googleapis.com/download/storage/v1/b/kaggle-user-content/o/inbox%2F551520%2Fb2bcd1d20c908ff1a6b1539f568f02a6%2FScreen%20Shot%202020-01-23%20at%2011.18.13%20AM.png?generation=1579778355998979&alt=media)\n", "728715": "Can you link us to the paper where this came from?", "729279": "paper link: https://staff.fnwi.uva.nl/t.e.j.mensink/publications/bolles17cvprwmf.pdf" } }