{ "id": 123981, "title": "Liveness detection ", "url": "/competitions/deepfake-detection-challenge/discussion/123981", "author_name": "B. Allabadi", "post_date": "2020-01-01T04:12:32.285000", "votes": 0, "comment_count": 0, "views": 0, "content": "

I noticed in some samples there are pictures of faces either on walls or shirts and these may be in all frames of the video.\n\"face_on_shirt\"\nI'm thinking that a good model may need to have liveness checking of some sort to avoid 2D faces from interfering with face detectors, since these faces may be artistic and may result in some videos being considered fake.

", "messages": [ { "id": 707564, "postDate": "2020-01-01T04:12:32.287Z", "content": "

I noticed in some samples there are pictures of faces either on walls or shirts and these may be in all frames of the video.\n\"face_on_shirt\"\nI'm thinking that a good model may need to have liveness checking of some sort to avoid 2D faces from interfering with face detectors, since these faces may be artistic and may result in some videos being considered fake.

", "rawMarkdown": "I noticed in some samples there are pictures of faces either on walls or shirts and these may be in all frames of the video.\n![face_on_shirt](https://www.kaggleusercontent.com/kf/25918383/eyJhbGciOiJkaXIiLCJlbmMiOiJBMTI4Q0JDLUhTMjU2In0..JXBCl6ehZ85Kz9i6wXleeA.-FpqLMOHd4DeyA6nZDkzhFEUKVTaMcUpQBTSZoLlRahlNplduiaUB_eB-ZMMsIm77LIi8rGL2v75gVaxUUUSygMW2_1huvNoxHCXpv0clwYIgn9QDeUEBvDOKBpD0A0CQ1ckXFq6hRScSm5opuzyui9RzlTg9pQ6XaHFCNie5mCVK8np1z8qcw3MoFSF7gJvE9-6Lc1e6hNAfGKA_eFB2A.2K1Ob5nrEKi6RJAsdeR1Kw/__results___files/__results___28_26.png)\nI'm thinking that a good model may need to have liveness checking of some sort to avoid 2D faces from interfering with face detectors, since these faces may be artistic and may result in some videos being considered fake.\n\n\n" } ], "comments": [], "raw_markdown_by_id": { "707564": "I noticed in some samples there are pictures of faces either on walls or shirts and these may be in all frames of the video.\n![face_on_shirt](https://www.kaggleusercontent.com/kf/25918383/eyJhbGciOiJkaXIiLCJlbmMiOiJBMTI4Q0JDLUhTMjU2In0..JXBCl6ehZ85Kz9i6wXleeA.-FpqLMOHd4DeyA6nZDkzhFEUKVTaMcUpQBTSZoLlRahlNplduiaUB_eB-ZMMsIm77LIi8rGL2v75gVaxUUUSygMW2_1huvNoxHCXpv0clwYIgn9QDeUEBvDOKBpD0A0CQ1ckXFq6hRScSm5opuzyui9RzlTg9pQ6XaHFCNie5mCVK8np1z8qcw3MoFSF7gJvE9-6Lc1e6hNAfGKA_eFB2A.2K1Ob5nrEKi6RJAsdeR1Kw/__results___files/__results___28_26.png)\nI'm thinking that a good model may need to have liveness checking of some sort to avoid 2D faces from interfering with face detectors, since these faces may be artistic and may result in some videos being considered fake.\n\n\n" } }