{
  "id": 128291,
  "title": "All the faces of the train sample",
  "url": "/competitions/deepfake-detection-challenge/discussion/128291",
  "author_name": "",
  "post_date": "2020-01-30T06:54:18.331650800Z",
  "votes": 7,
  "comment_count": 9,
  "views": 0,
  "content": "<p>Hi everyone.</p>\n\n<p>I have upload a database of all the faces of the train sample videos.\n<a href=\"https://www.kaggle.com/itamargr/dfdc-faces-of-the-train-sample/\">https://www.kaggle.com/itamargr/dfdc-faces-of-the-train-sample/</a></p>\n\n<p>Feel free to use it in your kernels</p>\n\n<p>Enjoy, Itamar</p>",
  "messages": [
    {
      "id": "732725",
      "postDate": "01/30/2020 06:54:18",
      "content": "<p>Hi everyone.</p>\n\n<p>I have upload a database of all the faces of the train sample videos.\n<a href=\"https://www.kaggle.com/itamargr/dfdc-faces-of-the-train-sample/\">https://www.kaggle.com/itamargr/dfdc-faces-of-the-train-sample/</a></p>\n\n<p>Feel free to use it in your kernels</p>\n\n<p>Enjoy, Itamar</p>",
      "rawMarkdown": "Hi everyone.\n\nI have upload a database of all the faces of the train sample videos.\nhttps://www.kaggle.com/itamargr/dfdc-faces-of-the-train-sample/\n\nFeel free to use it in your kernels\n\nEnjoy, Itamar",
      "votes": null
    },
    {
      "id": "732738",
      "postDate": "01/30/2020 07:28:56",
      "content": "<p>Does this mean the train sample videos or the entire train videos?</p>",
      "rawMarkdown": "Does this mean the train sample videos or the entire train videos?",
      "votes": null
    },
    {
      "id": "732792",
      "postDate": "01/30/2020 09:15:37",
      "content": "<p>Thanks for the share - pretty sure I will have questions once I download and look at some images.  But the first was how you decided train vs validation?</p>",
      "rawMarkdown": "Thanks for the share - pretty sure I will have questions once I download and look at some images.  But the first was how you decided train vs validation?",
      "votes": null
    },
    {
      "id": "732843",
      "postDate": "01/30/2020 10:54:22",
      "content": "<p>It is only the train sample videos.\nThe faces of the entire train videos dataset are too big to be entered in Kaggle</p>",
      "rawMarkdown": "It is only the train sample videos.\nThe faces of the entire train videos dataset are too big to be entered in Kaggle",
      "votes": null
    },
    {
      "id": "732844",
      "postDate": "01/30/2020 10:56:00",
      "content": "<p>I choose clips randomly. But I made sure that the real/fake ratio of the clips will be identical in the train and the validation set</p>",
      "rawMarkdown": "I choose clips randomly. But I made sure that the real/fake ratio of the clips will be identical in the train and the validation set",
      "votes": null
    },
    {
      "id": "733596",
      "postDate": "01/31/2020 10:29:25",
      "content": "<p>I think you'd be better off splitting by either the original video used or even the chunk name. Unless you have and I've misunderstood.</p>\n\n<p>There is still a leak with actors being present across chunks, but you don't want the same base videos to be present across train and validation.</p>",
      "rawMarkdown": "I think you'd be better off splitting by either the original video used or even the chunk name. Unless you have and I've misunderstood.\n\nThere is still a leak with actors being present across chunks, but you don't want the same base videos to be present across train and validation.",
      "votes": null
    },
    {
      "id": "734053",
      "postDate": "01/31/2020 21:45:18",
      "content": "<p>thanks for sharing</p>",
      "rawMarkdown": "thanks for sharing",
      "votes": null
    },
    {
      "id": "734666",
      "postDate": "02/01/2020 20:03:06",
      "content": "<p>You are right. I have splitted the clips into train and validation sts and than I extracted the images from them. So there is no base clip that is splitted between train and validation</p>",
      "rawMarkdown": "You are right. I have splitted the clips into train and validation sts and than I extracted the images from them. So there is no base clip that is splitted between train and validation",
      "votes": null
    },
    {
      "id": "734798",
      "postDate": "02/02/2020 02:37:41",
      "content": "<p>Thanks for sharing</p>",
      "rawMarkdown": "Thanks for sharing",
      "votes": null
    },
    {
      "id": "734821",
      "postDate": "02/02/2020 03:42:57",
      "content": "<p>Thank you very much.</p>",
      "rawMarkdown": "Thank you very much.",
      "votes": null
    }
  ],
  "comments": [
    {
      "id": 732738,
      "author_name": "chenshen03",
      "author_url": "",
      "post_date": "01/30/2020 07:28:56",
      "content": "<p>Does this mean the train sample videos or the entire train videos?</p>",
      "votes": null,
      "replies": [
        {
          "id": 732843,
          "author_name": "itamargr",
          "author_url": "",
          "post_date": "01/30/2020 10:54:22",
          "content": "<p>It is only the train sample videos.\nThe faces of the entire train videos dataset are too big to be entered in Kaggle</p>",
          "votes": null,
          "replies": []
        },
        {
          "id": 734821,
          "author_name": "chenshen03",
          "author_url": "",
          "post_date": "02/02/2020 03:42:57",
          "content": "<p>Thank you very much.</p>",
          "votes": null,
          "replies": []
        }
      ]
    },
    {
      "id": 732792,
      "author_name": "pcjimmmy",
      "author_url": "",
      "post_date": "01/30/2020 09:15:37",
      "content": "<p>Thanks for the share - pretty sure I will have questions once I download and look at some images.  But the first was how you decided train vs validation?</p>",
      "votes": null,
      "replies": [
        {
          "id": 732844,
          "author_name": "itamargr",
          "author_url": "",
          "post_date": "01/30/2020 10:56:00",
          "content": "<p>I choose clips randomly. But I made sure that the real/fake ratio of the clips will be identical in the train and the validation set</p>",
          "votes": null,
          "replies": []
        },
        {
          "id": 733596,
          "author_name": "jamesphoward",
          "author_url": "",
          "post_date": "01/31/2020 10:29:25",
          "content": "<p>I think you'd be better off splitting by either the original video used or even the chunk name. Unless you have and I've misunderstood.</p>\n\n<p>There is still a leak with actors being present across chunks, but you don't want the same base videos to be present across train and validation.</p>",
          "votes": null,
          "replies": []
        },
        {
          "id": 734666,
          "author_name": "itamargr",
          "author_url": "",
          "post_date": "02/01/2020 20:03:06",
          "content": "<p>You are right. I have splitted the clips into train and validation sts and than I extracted the images from them. So there is no base clip that is splitted between train and validation</p>",
          "votes": null,
          "replies": []
        }
      ]
    },
    {
      "id": 734053,
      "author_name": "kaggleurroad",
      "author_url": "",
      "post_date": "01/31/2020 21:45:18",
      "content": "<p>thanks for sharing</p>",
      "votes": null,
      "replies": []
    },
    {
      "id": 734798,
      "author_name": "ajayprakashnair",
      "author_url": "",
      "post_date": "02/02/2020 02:37:41",
      "content": "<p>Thanks for sharing</p>",
      "votes": null,
      "replies": []
    }
  ],
  "raw_markdown_by_id": {
    "732725": "Hi everyone.\n\nI have upload a database of all the faces of the train sample videos.\nhttps://www.kaggle.com/itamargr/dfdc-faces-of-the-train-sample/\n\nFeel free to use it in your kernels\n\nEnjoy, Itamar",
    "732738": "Does this mean the train sample videos or the entire train videos?",
    "732792": "Thanks for the share - pretty sure I will have questions once I download and look at some images.  But the first was how you decided train vs validation?",
    "732843": "It is only the train sample videos.\nThe faces of the entire train videos dataset are too big to be entered in Kaggle",
    "732844": "I choose clips randomly. But I made sure that the real/fake ratio of the clips will be identical in the train and the validation set",
    "733596": "I think you'd be better off splitting by either the original video used or even the chunk name. Unless you have and I've misunderstood.\n\nThere is still a leak with actors being present across chunks, but you don't want the same base videos to be present across train and validation.",
    "734053": "thanks for sharing",
    "734666": "You are right. I have splitted the clips into train and validation sts and than I extracted the images from them. So there is no base clip that is splitted between train and validation",
    "734798": "Thanks for sharing",
    "734821": "Thank you very much."
  },
  "source": "meta"
}