{
  "id": 149150,
  "title": "Tiles training with tensorflow ",
  "url": "/competitions/prostate-cancer-grade-assessment/discussion/149150",
  "author_name": "Tsai29",
  "post_date": "2020-05-07T04:01:50.302000",
  "votes": 6,
  "comment_count": 0,
  "views": 0,
  "content": "<p>Hi guys, I had created a public kernel about how to train the tile images like this <a href=\"https://www.kaggle.com/iafoss/panda-concat-tile-pooling-starter-0-79-lb\">kernel</a> on tensorflow framework. </p>\n\n<p><a href=\"https://www.kaggle.com/xiejialun/panda-tiles-training-on-tensorflow-0-7-cv\">https://www.kaggle.com/xiejialun/panda-tiles-training-on-tensorflow-0-7-cv</a></p>\n\n<p>I believe it can reach the similar performance as in fast.ai/pytorch framework after some fine tune.\nJust want to share some good start points for who only use tensorflow framework.\nGood luck :)</p>\n\n<p>By the way, is there anyone also still trying to solve the task with segmentation model?\nI would like to hear some performance sharing\nMy best segmentation performance now:</p>\n\n<p>efficientnetb0 backbone unet, single fold\nCV : 0.64\nLB : 0.66</p>",
  "messages": [
    {
      "id": 836549,
      "postDate": "2020-05-07T04:01:50.303Z",
      "content": "<p>Hi guys, I had created a public kernel about how to train the tile images like this <a href=\"https://www.kaggle.com/iafoss/panda-concat-tile-pooling-starter-0-79-lb\">kernel</a> on tensorflow framework. </p>\n\n<p><a href=\"https://www.kaggle.com/xiejialun/panda-tiles-training-on-tensorflow-0-7-cv\">https://www.kaggle.com/xiejialun/panda-tiles-training-on-tensorflow-0-7-cv</a></p>\n\n<p>I believe it can reach the similar performance as in fast.ai/pytorch framework after some fine tune.\nJust want to share some good start points for who only use tensorflow framework.\nGood luck :)</p>\n\n<p>By the way, is there anyone also still trying to solve the task with segmentation model?\nI would like to hear some performance sharing\nMy best segmentation performance now:</p>\n\n<p>efficientnetb0 backbone unet, single fold\nCV : 0.64\nLB : 0.66</p>",
      "rawMarkdown": "Hi guys, I had created a public kernel about how to train the tile images like this [kernel](https://www.kaggle.com/iafoss/panda-concat-tile-pooling-starter-0-79-lb) on tensorflow framework. \n\nhttps://www.kaggle.com/xiejialun/panda-tiles-training-on-tensorflow-0-7-cv\n\nI believe it can reach the similar performance as in fast.ai/pytorch framework after some fine tune.\nJust want to share some good start points for who only use tensorflow framework.\nGood luck :)\n\nBy the way, is there anyone also still trying to solve the task with segmentation model?\nI would like to hear some performance sharing\nMy best segmentation performance now:\n\nefficientnetb0 backbone unet, single fold\nCV : 0.64\nLB : 0.66",
      "votes": 6
    }
  ],
  "comments": [],
  "raw_markdown_by_id": {
    "836549": "Hi guys, I had created a public kernel about how to train the tile images like this [kernel](https://www.kaggle.com/iafoss/panda-concat-tile-pooling-starter-0-79-lb) on tensorflow framework. \n\nhttps://www.kaggle.com/xiejialun/panda-tiles-training-on-tensorflow-0-7-cv\n\nI believe it can reach the similar performance as in fast.ai/pytorch framework after some fine tune.\nJust want to share some good start points for who only use tensorflow framework.\nGood luck :)\n\nBy the way, is there anyone also still trying to solve the task with segmentation model?\nI would like to hear some performance sharing\nMy best segmentation performance now:\n\nefficientnetb0 backbone unet, single fold\nCV : 0.64\nLB : 0.66"
  }
}