{
  "id": 359371,
  "title": "Silver medal solution: ViT-L/14 with test time augmentations",
  "content": "<p>Thank you to the host, organizers and all participants!</p>\n<p>You can check out my solution here: <a href=\"https://www.kaggle.com/code/slavabarkov/google-universal-image-embedding-submission\" target=\"_blank\">https://www.kaggle.com/code/slavabarkov/google-universal-image-embedding-submission</a></p>\n<p>Brief summary of my solution:</p>\n<ul>\n<li>ViT-L/14 encoder from the OpenCLIP implementation pre-trained on LAION-400M dataset, achieves 0.519 public/0.537 private if the input image is resized while maintaining the aspect ratio</li>\n<li>Projection layer trained with ArcFace on the Products-10k dataset based on <a href=\"https://www.kaggle.com/motono0223\" target=\"_blank\">@motono0223</a> work, achieves 0.598 public/0.608 private</li>\n<li>Multiple test time augmentations, achieves 0.628 public/0.631 private: <ul>\n<li>Predefined crops - center crop and two crops with preselected offset</li>\n<li>Pixel dropout - drops pixels from image according to predefined mask</li></ul></li>\n</ul>",
  "comments": [],
  "comment_count": 0,
  "messages": [
    {
      "id": 1983036,
      "postDate": "2022-10-11T18:41:51.813Z",
      "content": "<p>Thank you to the host, organizers and all participants!</p>\n<p>You can check out my solution here: <a href=\"https://www.kaggle.com/code/slavabarkov/google-universal-image-embedding-submission\" target=\"_blank\">https://www.kaggle.com/code/slavabarkov/google-universal-image-embedding-submission</a></p>\n<p>Brief summary of my solution:</p>\n<ul>\n<li>ViT-L/14 encoder from the OpenCLIP implementation pre-trained on LAION-400M dataset, achieves 0.519 public/0.537 private if the input image is resized while maintaining the aspect ratio</li>\n<li>Projection layer trained with ArcFace on the Products-10k dataset based on <a href=\"https://www.kaggle.com/motono0223\" target=\"_blank\">@motono0223</a> work, achieves 0.598 public/0.608 private</li>\n<li>Multiple test time augmentations, achieves 0.628 public/0.631 private: <ul>\n<li>Predefined crops - center crop and two crops with preselected offset</li>\n<li>Pixel dropout - drops pixels from image according to predefined mask</li></ul></li>\n</ul>",
      "rawMarkdown": "Thank you to the host, organizers and all participants!\n\nYou can check out my solution here: https://www.kaggle.com/code/slavabarkov/google-universal-image-embedding-submission\n\nBrief summary of my solution:\n- ViT-L/14 encoder from the OpenCLIP implementation pre-trained on LAION-400M dataset, achieves 0.519 public/0.537 private if the input image is resized while maintaining the aspect ratio\n- Projection layer trained with ArcFace on the Products-10k dataset based on @motono0223 work, achieves 0.598 public/0.608 private\n- Multiple test time augmentations, achieves 0.628 public/0.631 private: \n\t- Predefined crops - center crop and two crops with preselected offset\n\t- Pixel dropout - drops pixels from image according to predefined mask",
      "votes": 13
    }
  ],
  "complete": true,
  "source": "live",
  "fetched_at": "2026-10-22T08:32:17Z",
  "competition_id": 36414,
  "competition_slug": "google-universal-image-embedding",
  "team_id": 8890179,
  "solution_url": "/competitions/google-universal-image-embedding/writeups/viacheslav-barkov-silver-medal-solution-vit-l-14-w"
}