{
  "id": 173735,
  "title": "Any luck with BiT (Big transfer)?",
  "url": "/competitions/siim-isic-melanoma-classification/discussion/173735",
  "author_name": "DimitreOliveira",
  "post_date": "2020-08-10T13:12:25.698000",
  "votes": 3,
  "comment_count": 0,
  "views": 0,
  "content": "<p>Hey guys, before joining this competition I saw some videos/articles about <a href=\"https://arxiv.org/pdf/1912.11370.pdf\" target=\"_blank\">big transfer (BiT)</a> and had high hopes for it on the competition, but I did not get it to work very well (at least not compared to EfficientNet), I saw that there are some notebooks using it but not with high CV/LB, is anyone getting good results with it?</p>\n<p>For those willing to give a try here is a <a href=\"https://colab.research.google.com/github/google-research/big_transfer/blob/master/colabs/big_transfer_tf2.ipynb#scrollTo=GXY0spPLdxkX\" target=\"_blank\">nice tutorial</a> and a <a href=\"https://www.kaggle.com/dimitreoliveira/big-transfer-bit-weights\" target=\"_blank\">dataset</a> with the weights.</p>",
  "messages": [
    {
      "id": 965203,
      "postDate": "2020-08-10T13:12:25.697Z",
      "content": "<p>Hey guys, before joining this competition I saw some videos/articles about <a href=\"https://arxiv.org/pdf/1912.11370.pdf\" target=\"_blank\">big transfer (BiT)</a> and had high hopes for it on the competition, but I did not get it to work very well (at least not compared to EfficientNet), I saw that there are some notebooks using it but not with high CV/LB, is anyone getting good results with it?</p>\n<p>For those willing to give a try here is a <a href=\"https://colab.research.google.com/github/google-research/big_transfer/blob/master/colabs/big_transfer_tf2.ipynb#scrollTo=GXY0spPLdxkX\" target=\"_blank\">nice tutorial</a> and a <a href=\"https://www.kaggle.com/dimitreoliveira/big-transfer-bit-weights\" target=\"_blank\">dataset</a> with the weights.</p>",
      "rawMarkdown": "Hey guys, before joining this competition I saw some videos/articles about [big transfer (BiT)](https://arxiv.org/pdf/1912.11370.pdf) and had high hopes for it on the competition, but I did not get it to work very well (at least not compared to EfficientNet), I saw that there are some notebooks using it but not with high CV/LB, is anyone getting good results with it?\n\nFor those willing to give a try here is a [nice tutorial](https://colab.research.google.com/github/google-research/big_transfer/blob/master/colabs/big_transfer_tf2.ipynb#scrollTo=GXY0spPLdxkX) and a [dataset](https://www.kaggle.com/dimitreoliveira/big-transfer-bit-weights) with the weights.",
      "votes": 3
    }
  ],
  "comments": [],
  "raw_markdown_by_id": {
    "965203": "Hey guys, before joining this competition I saw some videos/articles about [big transfer (BiT)](https://arxiv.org/pdf/1912.11370.pdf) and had high hopes for it on the competition, but I did not get it to work very well (at least not compared to EfficientNet), I saw that there are some notebooks using it but not with high CV/LB, is anyone getting good results with it?\n\nFor those willing to give a try here is a [nice tutorial](https://colab.research.google.com/github/google-research/big_transfer/blob/master/colabs/big_transfer_tf2.ipynb#scrollTo=GXY0spPLdxkX) and a [dataset](https://www.kaggle.com/dimitreoliveira/big-transfer-bit-weights) with the weights."
  }
}