{
  "id": 129470,
  "title": "Identify fake videos by people's movements",
  "url": "/competitions/deepfake-detection-challenge/discussion/129470",
  "author_name": "",
  "post_date": "2020-02-08T08:20:18.183266100Z",
  "votes": 4,
  "comment_count": 2,
  "views": 0,
  "content": "<p>Is it possible to identify fake videos on the movements of people and their parts of body or face?</p>\n\n<p>When the synthesis of human movements is performed on the basis of the images themselves, rather than human dynamic models, unnatural movements with a torn rhythm sometimes occur. \nApplying this approach speeds up identification or improves accuracy.</p>\n\n<p>Did anyone find any similar patterns?</p>",
  "messages": [
    {
      "id": "739701",
      "postDate": "02/08/2020 08:20:18",
      "content": "<p>Is it possible to identify fake videos on the movements of people and their parts of body or face?</p>\n\n<p>When the synthesis of human movements is performed on the basis of the images themselves, rather than human dynamic models, unnatural movements with a torn rhythm sometimes occur. \nApplying this approach speeds up identification or improves accuracy.</p>\n\n<p>Did anyone find any similar patterns?</p>",
      "rawMarkdown": "Is it possible to identify fake videos on the movements of people and their parts of body or face?\n \nWhen the synthesis of human movements is performed on the basis of the images themselves, rather than human dynamic models, unnatural movements with a torn rhythm sometimes occur. \nApplying this approach speeds up identification or improves accuracy.\n\nDid anyone find any similar patterns?",
      "votes": null
    },
    {
      "id": "739771",
      "postDate": "02/08/2020 10:47:44",
      "content": "<p>However, general  methods of dealing with such kind of movements always regard these artifacts as noise (maybe generated by the photographers).\nAny solution to maximize the unnatural movements?</p>",
      "rawMarkdown": "However, general  methods of dealing with such kind of movements always regard these artifacts as noise (maybe generated by the photographers).\nAny solution to maximize the unnatural movements?",
      "votes": null
    },
    {
      "id": "740691",
      "postDate": "02/09/2020 16:58:39",
      "content": "<p>Yes. This might work by subtracting the current frame to the previous frame. Humans in general will not do such big movements in one frame. But in deep fake, the movement will be quite unnatural. HOWEVER, I tried this before but ended up got disturbed by as <a href=\"/kunlin612\">@kunlin612</a> mentioned noise. The two frame looks exactly the same BUT there is still a lot of difference somehow.</p>",
      "rawMarkdown": "Yes. This might work by subtracting the current frame to the previous frame. Humans in general will not do such big movements in one frame. But in deep fake, the movement will be quite unnatural. HOWEVER, I tried this before but ended up got disturbed by as @kunlin612 mentioned noise. The two frame looks exactly the same BUT there is still a lot of difference somehow.",
      "votes": null
    }
  ],
  "comments": [
    {
      "id": 739771,
      "author_name": "kunlin612",
      "author_url": "",
      "post_date": "02/08/2020 10:47:44",
      "content": "<p>However, general  methods of dealing with such kind of movements always regard these artifacts as noise (maybe generated by the photographers).\nAny solution to maximize the unnatural movements?</p>",
      "votes": null,
      "replies": []
    },
    {
      "id": 740691,
      "author_name": "unkownhihi",
      "author_url": "",
      "post_date": "02/09/2020 16:58:39",
      "content": "<p>Yes. This might work by subtracting the current frame to the previous frame. Humans in general will not do such big movements in one frame. But in deep fake, the movement will be quite unnatural. HOWEVER, I tried this before but ended up got disturbed by as <a href=\"/kunlin612\">@kunlin612</a> mentioned noise. The two frame looks exactly the same BUT there is still a lot of difference somehow.</p>",
      "votes": null,
      "replies": []
    }
  ],
  "raw_markdown_by_id": {
    "739701": "Is it possible to identify fake videos on the movements of people and their parts of body or face?\n \nWhen the synthesis of human movements is performed on the basis of the images themselves, rather than human dynamic models, unnatural movements with a torn rhythm sometimes occur. \nApplying this approach speeds up identification or improves accuracy.\n\nDid anyone find any similar patterns?",
    "739771": "However, general  methods of dealing with such kind of movements always regard these artifacts as noise (maybe generated by the photographers).\nAny solution to maximize the unnatural movements?",
    "740691": "Yes. This might work by subtracting the current frame to the previous frame. Humans in general will not do such big movements in one frame. But in deep fake, the movement will be quite unnatural. HOWEVER, I tried this before but ended up got disturbed by as @kunlin612 mentioned noise. The two frame looks exactly the same BUT there is still a lot of difference somehow."
  },
  "source": "meta"
}