{
  "id": 230216,
  "title": "EfficientNetV2: Smaller Models and Faster Training",
  "url": "/competitions/hpa-single-cell-image-classification/discussion/230216",
  "author_name": "Saurabh Shahane",
  "post_date": "2021-04-02T15:47:33.196000",
  "votes": 13,
  "comment_count": 3,
  "views": 0,
  "content": "<p>With progressive learning, EfficientNetV2 significantly outperforms previous models on ImageNet, including ViT by 2.0% acc. while training 5x-11x faster.</p>\n<p>Paper - <a href=\"https://arxiv.org/pdf/2104.00298.pdf\" target=\"_blank\">https://arxiv.org/pdf/2104.00298.pdf</a><br>\nCode - <a href=\"https://github.com/google/automl\" target=\"_blank\">https://github.com/google/automl</a></p>\n<p><img src=\"https://pbs.twimg.com/media/Ex7ila1U8AMMwO0?format=jpg&amp;name=large\" alt=\"img\"></p>",
  "messages": [
    {
      "id": 1261002,
      "postDate": "2021-04-02T15:47:33.197Z",
      "content": "<p>With progressive learning, EfficientNetV2 significantly outperforms previous models on ImageNet, including ViT by 2.0% acc. while training 5x-11x faster.</p>\n<p>Paper - <a href=\"https://arxiv.org/pdf/2104.00298.pdf\" target=\"_blank\">https://arxiv.org/pdf/2104.00298.pdf</a><br>\nCode - <a href=\"https://github.com/google/automl\" target=\"_blank\">https://github.com/google/automl</a></p>\n<p><img src=\"https://pbs.twimg.com/media/Ex7ila1U8AMMwO0?format=jpg&amp;name=large\" alt=\"img\"></p>",
      "rawMarkdown": "With progressive learning, EfficientNetV2 significantly outperforms previous models on ImageNet, including ViT by 2.0% acc. while training 5x-11x faster.\n\nPaper - https://arxiv.org/pdf/2104.00298.pdf\nCode - https://github.com/google/automl\n\n![img](https://pbs.twimg.com/media/Ex7ila1U8AMMwO0?format=jpg&name=large)",
      "votes": 14
    },
    {
      "id": 1277342,
      "postDate": "2021-04-18T16:29:16.377Z",
      "content": "<p>place holder for code!!!</p>",
      "rawMarkdown": "place holder for code!!!",
      "votes": 1,
      "replies": [
        {
          "id": 1281929,
          "postDate": "2021-04-23T12:55:58.343Z",
          "content": "<p><a href=\"https://www.kaggle.com/dragonzhang\" target=\"_blank\">@dragonzhang</a> lol Don't be sad. Let's code it out!</p>",
          "rawMarkdown": "@dragonzhang lol Don't be sad. Let's code it out!",
          "votes": 1
        }
      ]
    },
    {
      "id": 1273181,
      "postDate": "2021-04-14T06:36:42.650Z",
      "content": "<p>There is a Meta Pseudo Labels is a champion on ImageNet now.</p>\n<p><a href=\"https://arxiv.org/pdf/2003.10580v4.pdf\" target=\"_blank\">https://arxiv.org/pdf/2003.10580v4.pdf</a><br>\n<a href=\"https://github.com/google-research/google-research/tree/master/meta_pseudo_labels\" target=\"_blank\">https://github.com/google-research/google-research/tree/master/meta_pseudo_labels</a></p>",
      "rawMarkdown": "There is a Meta Pseudo Labels is a champion on ImageNet now.\n\nhttps://arxiv.org/pdf/2003.10580v4.pdf\nhttps://github.com/google-research/google-research/tree/master/meta_pseudo_labels",
      "votes": 1
    }
  ],
  "comments": [
    {
      "id": 1277342,
      "author_name": "dragon zhang",
      "author_url": "",
      "post_date": "2021-04-18T16:29:16.377000",
      "content": "<p>place holder for code!!!</p>",
      "votes": 1,
      "replies": [
        {
          "id": 1281929,
          "author_name": "Chien-Hsiang Hung",
          "author_url": "",
          "post_date": "2021-04-23T12:55:58.343000",
          "content": "<p><a href=\"https://www.kaggle.com/dragonzhang\" target=\"_blank\">@dragonzhang</a> lol Don't be sad. Let's code it out!</p>",
          "votes": 1,
          "replies": []
        }
      ]
    },
    {
      "id": 1273181,
      "author_name": "ZavodRobotov",
      "author_url": "",
      "post_date": "2021-04-14T06:36:42.650000",
      "content": "<p>There is a Meta Pseudo Labels is a champion on ImageNet now.</p>\n<p><a href=\"https://arxiv.org/pdf/2003.10580v4.pdf\" target=\"_blank\">https://arxiv.org/pdf/2003.10580v4.pdf</a><br>\n<a href=\"https://github.com/google-research/google-research/tree/master/meta_pseudo_labels\" target=\"_blank\">https://github.com/google-research/google-research/tree/master/meta_pseudo_labels</a></p>",
      "votes": 1,
      "replies": []
    }
  ],
  "raw_markdown_by_id": {
    "1261002": "With progressive learning, EfficientNetV2 significantly outperforms previous models on ImageNet, including ViT by 2.0% acc. while training 5x-11x faster.\n\nPaper - https://arxiv.org/pdf/2104.00298.pdf\nCode - https://github.com/google/automl\n\n![img](https://pbs.twimg.com/media/Ex7ila1U8AMMwO0?format=jpg&name=large)",
    "1277342": "place holder for code!!!",
    "1273181": "There is a Meta Pseudo Labels is a champion on ImageNet now.\n\nhttps://arxiv.org/pdf/2003.10580v4.pdf\nhttps://github.com/google-research/google-research/tree/master/meta_pseudo_labels"
  }
}