{
  "id": 128069,
  "title": "Frame 150 from ALL Videos",
  "url": "/competitions/deepfake-detection-challenge/discussion/128069",
  "author_name": "",
  "post_date": "2020-01-28T17:36:24.570814800Z",
  "votes": 6,
  "comment_count": 8,
  "views": 0,
  "content": "<p>94,147 images pulled from DFDC dataset. \n77,944 fake and 16,203 real images\nCropped 224x224</p>\n\n<p><a href=\"https://www.kaggle.com/sciarrilli/dfdc-f150\">https://www.kaggle.com/sciarrilli/dfdc-f150</a></p>",
  "messages": [
    {
      "id": "731491",
      "postDate": "01/28/2020 17:36:24",
      "content": "<p>94,147 images pulled from DFDC dataset. \n77,944 fake and 16,203 real images\nCropped 224x224</p>\n\n<p><a href=\"https://www.kaggle.com/sciarrilli/dfdc-f150\">https://www.kaggle.com/sciarrilli/dfdc-f150</a></p>",
      "rawMarkdown": "94,147 images pulled from DFDC dataset. \n77,944 fake and 16,203 real images\nCropped 224x224\n\nhttps://www.kaggle.com/sciarrilli/dfdc-f150",
      "votes": null
    },
    {
      "id": "731507",
      "postDate": "01/28/2020 18:14:08",
      "content": "<p>You added some type of filter? They look blue</p>",
      "rawMarkdown": "You added some type of filter? They look blue",
      "votes": null
    },
    {
      "id": "731521",
      "postDate": "01/28/2020 18:36:20",
      "content": "<p>yes they were converted to gray scale. </p>\n\n<p>gray = cv2.cvtColor(image, cv2.COLOR_BGR2GRAY)</p>",
      "rawMarkdown": "yes they were converted to gray scale. \n\ngray = cv2.cvtColor(image, cv2.COLOR_BGR2GRAY)",
      "votes": null
    },
    {
      "id": "731523",
      "postDate": "01/28/2020 18:39:28",
      "content": "<p>I could be wrong - I think color will play a role - slightly off shades could feature deepfake</p>",
      "rawMarkdown": "I could be wrong - I think color will play a role - slightly off shades could feature deepfake",
      "votes": null
    },
    {
      "id": "731534",
      "postDate": "01/28/2020 19:04:56",
      "content": "<p>Sorry its actually normalization for resnet</p>\n\n<p>transform = transforms.Compose([transforms.Resize([224, 224]),\n                                transforms.ToTensor(),\n                                transforms.Normalize(mean=[0.485, 0.456, 0.406], std=[0.229, 0.224, 0.225])])</p>",
      "rawMarkdown": "Sorry its actually normalization for resnet\n\ntransform = transforms.Compose([transforms.Resize([224, 224]),\n                                transforms.ToTensor(),\n                                transforms.Normalize(mean=[0.485, 0.456, 0.406], std=[0.229, 0.224, 0.225])])",
      "votes": null
    },
    {
      "id": "731724",
      "postDate": "01/29/2020 00:58:05",
      "content": "<p>You might forgot to convert it from BGR to RGB. They look blue and that is what BGR images look like.</p>",
      "rawMarkdown": "You might forgot to convert it from BGR to RGB. They look blue and that is what BGR images look like.",
      "votes": null
    },
    {
      "id": "733332",
      "postDate": "01/31/2020 01:27:25",
      "content": "<p>While this is useful, one thing to note is that some videos are not modified for every single frame. If a person turns to face away from the camera at any time then no modifications are made in those frames.</p>",
      "rawMarkdown": "While this is useful, one thing to note is that some videos are not modified for every single frame. If a person turns to face away from the camera at any time then no modifications are made in those frames.",
      "votes": null
    },
    {
      "id": "733335",
      "postDate": "01/31/2020 01:38:38",
      "content": "<p>understood</p>",
      "rawMarkdown": "understood",
      "votes": null
    },
    {
      "id": "733577",
      "postDate": "01/31/2020 10:13:37",
      "content": "<p>The images were saved on BGR format. </p>",
      "rawMarkdown": "The images were saved on BGR format.",
      "votes": null
    }
  ],
  "comments": [
    {
      "id": 731507,
      "author_name": "sethkitchen",
      "author_url": "",
      "post_date": "01/28/2020 18:14:08",
      "content": "<p>You added some type of filter? They look blue</p>",
      "votes": null,
      "replies": [
        {
          "id": 733577,
          "author_name": "luisgoncalo",
          "author_url": "",
          "post_date": "01/31/2020 10:13:37",
          "content": "<p>The images were saved on BGR format. </p>",
          "votes": null,
          "replies": []
        }
      ]
    },
    {
      "id": 731521,
      "author_name": "sciarrilli",
      "author_url": "",
      "post_date": "01/28/2020 18:36:20",
      "content": "<p>yes they were converted to gray scale. </p>\n\n<p>gray = cv2.cvtColor(image, cv2.COLOR_BGR2GRAY)</p>",
      "votes": null,
      "replies": [
        {
          "id": 731523,
          "author_name": "sethkitchen",
          "author_url": "",
          "post_date": "01/28/2020 18:39:28",
          "content": "<p>I could be wrong - I think color will play a role - slightly off shades could feature deepfake</p>",
          "votes": null,
          "replies": []
        },
        {
          "id": 731534,
          "author_name": "sciarrilli",
          "author_url": "",
          "post_date": "01/28/2020 19:04:56",
          "content": "<p>Sorry its actually normalization for resnet</p>\n\n<p>transform = transforms.Compose([transforms.Resize([224, 224]),\n                                transforms.ToTensor(),\n                                transforms.Normalize(mean=[0.485, 0.456, 0.406], std=[0.229, 0.224, 0.225])])</p>",
          "votes": null,
          "replies": []
        }
      ]
    },
    {
      "id": 731724,
      "author_name": "unkownhihi",
      "author_url": "",
      "post_date": "01/29/2020 00:58:05",
      "content": "<p>You might forgot to convert it from BGR to RGB. They look blue and that is what BGR images look like.</p>",
      "votes": null,
      "replies": []
    },
    {
      "id": 733332,
      "author_name": "ldm314",
      "author_url": "",
      "post_date": "01/31/2020 01:27:25",
      "content": "<p>While this is useful, one thing to note is that some videos are not modified for every single frame. If a person turns to face away from the camera at any time then no modifications are made in those frames.</p>",
      "votes": null,
      "replies": [
        {
          "id": 733335,
          "author_name": "sciarrilli",
          "author_url": "",
          "post_date": "01/31/2020 01:38:38",
          "content": "<p>understood</p>",
          "votes": null,
          "replies": []
        }
      ]
    }
  ],
  "raw_markdown_by_id": {
    "731491": "94,147 images pulled from DFDC dataset. \n77,944 fake and 16,203 real images\nCropped 224x224\n\nhttps://www.kaggle.com/sciarrilli/dfdc-f150",
    "731507": "You added some type of filter? They look blue",
    "731521": "yes they were converted to gray scale. \n\ngray = cv2.cvtColor(image, cv2.COLOR_BGR2GRAY)",
    "731523": "I could be wrong - I think color will play a role - slightly off shades could feature deepfake",
    "731534": "Sorry its actually normalization for resnet\n\ntransform = transforms.Compose([transforms.Resize([224, 224]),\n                                transforms.ToTensor(),\n                                transforms.Normalize(mean=[0.485, 0.456, 0.406], std=[0.229, 0.224, 0.225])])",
    "731724": "You might forgot to convert it from BGR to RGB. They look blue and that is what BGR images look like.",
    "733332": "While this is useful, one thing to note is that some videos are not modified for every single frame. If a person turns to face away from the camera at any time then no modifications are made in those frames.",
    "733335": "understood",
    "733577": "The images were saved on BGR format."
  },
  "source": "meta"
}