{
  "id": 122013,
  "title": "Public Test Set videos are not the same as Training Set",
  "url": "/competitions/deepfake-detection-challenge/discussion/122013",
  "author_name": "Behrouz",
  "post_date": "2019-12-17T04:19:42.548000",
  "votes": 52,
  "comment_count": 0,
  "views": 0,
  "content": "<p>Based on this paper <a href=\"https://arxiv.org/abs/1910.08854\">The Deepfake Detection Challenge (DFDC) Preview Dataset</a> Test Sets contain 3 types of augmentations that we should generate to train our model</p>\n\n<p>(1) reduce the FPS of the video to 15; \n(2) reduce the resolution of the video to 1/4 of its original size;\n(3) reduce the overall encoding quality.</p>\n\n<p>The training set were left at their original resolution and quality, so deriving appropriate augmentations of the training set is left as an exercise to the researcher.</p>\n\n<p>No video was subjected to more than one augmentation.</p>",
  "messages": [
    {
      "id": 696789,
      "postDate": "2019-12-17T04:19:42.550Z",
      "content": "<p>Based on this paper <a href=\"https://arxiv.org/abs/1910.08854\">The Deepfake Detection Challenge (DFDC) Preview Dataset</a> Test Sets contain 3 types of augmentations that we should generate to train our model</p>\n\n<p>(1) reduce the FPS of the video to 15; \n(2) reduce the resolution of the video to 1/4 of its original size;\n(3) reduce the overall encoding quality.</p>\n\n<p>The training set were left at their original resolution and quality, so deriving appropriate augmentations of the training set is left as an exercise to the researcher.</p>\n\n<p>No video was subjected to more than one augmentation.</p>",
      "rawMarkdown": "Based on this paper [The Deepfake Detection Challenge (DFDC) Preview Dataset](https://arxiv.org/abs/1910.08854) Test Sets contain 3 types of augmentations that we should generate to train our model\n\n(1) reduce the FPS of the video to 15; \n(2) reduce the resolution of the video to 1/4 of its original size;\n(3) reduce the overall encoding quality.\n\nThe training set were left at their original resolution and quality, so deriving appropriate augmentations of the training set is left as an exercise to the researcher.\n\nNo video was subjected to more than one augmentation.\n",
      "votes": 50
    }
  ],
  "comments": [],
  "raw_markdown_by_id": {
    "696789": "Based on this paper [The Deepfake Detection Challenge (DFDC) Preview Dataset](https://arxiv.org/abs/1910.08854) Test Sets contain 3 types of augmentations that we should generate to train our model\n\n(1) reduce the FPS of the video to 15; \n(2) reduce the resolution of the video to 1/4 of its original size;\n(3) reduce the overall encoding quality.\n\nThe training set were left at their original resolution and quality, so deriving appropriate augmentations of the training set is left as an exercise to the researcher.\n\nNo video was subjected to more than one augmentation.\n"
  }
}