{
  "id": 475169,
  "title": "320 to 88th rank  would have reached Rank 22",
  "url": "/competitions/blood-vessel-segmentation/discussion/475169",
  "author_name": "Arunodhayan",
  "post_date": "2024-02-07T11:45:31.638000",
  "votes": 1,
  "comment_count": 0,
  "views": 0,
  "content": "<h2>preprocessing</h2>\n<ul>\n<li>Used the original Resolution for training by padding it with Zeros</li>\n<li>kidney 1 + 3 sparse for training </li>\n<li>kidney 3 for validation </li>\n</ul>\n<h3>DataAugmentation</h3>\n<ul>\n<li>Flip (Horizontal and Vertical)</li>\n<li>GAussian noise</li>\n<li>gamma</li>\n<li>scale rotate </li>\n</ul>\n<h2>Model</h2>\n<ul>\n<li>tu-resnet50d UNET</li>\n<li>Effcientnet b0 UNET</li>\n</ul>\n<p>Best Performance </p>\n<ul>\n<li>Public 0.805 Private 0.534 ensemble tu-resnet50D+ Efficientnet b0</li>\n<li>Public 0.696 Private 0.610  tu-resnet50D</li>\n</ul>",
  "messages": [
    {
      "id": 2641275,
      "postDate": "2024-02-07T11:45:31.640Z",
      "content": "<h2>preprocessing</h2>\n<ul>\n<li>Used the original Resolution for training by padding it with Zeros</li>\n<li>kidney 1 + 3 sparse for training </li>\n<li>kidney 3 for validation </li>\n</ul>\n<h3>DataAugmentation</h3>\n<ul>\n<li>Flip (Horizontal and Vertical)</li>\n<li>GAussian noise</li>\n<li>gamma</li>\n<li>scale rotate </li>\n</ul>\n<h2>Model</h2>\n<ul>\n<li>tu-resnet50d UNET</li>\n<li>Effcientnet b0 UNET</li>\n</ul>\n<p>Best Performance </p>\n<ul>\n<li>Public 0.805 Private 0.534 ensemble tu-resnet50D+ Efficientnet b0</li>\n<li>Public 0.696 Private 0.610  tu-resnet50D</li>\n</ul>",
      "rawMarkdown": "##preprocessing \n- Used the original Resolution for training by padding it with Zeros\n- kidney 1 + 3 sparse for training \n- kidney 3 for validation \n\n###DataAugmentation\n- Flip (Horizontal and Vertical)\n- GAussian noise\n- gamma\n- scale rotate \n\n##Model\n-  tu-resnet50d UNET\n-  Effcientnet b0 UNET\n\nBest Performance \n- Public 0.805 Private 0.534 ensemble tu-resnet50D+ Efficientnet b0\n- Public 0.696 Private 0.610  tu-resnet50D\n\n",
      "votes": 1
    }
  ],
  "comments": [],
  "raw_markdown_by_id": {
    "2641275": "##preprocessing \n- Used the original Resolution for training by padding it with Zeros\n- kidney 1 + 3 sparse for training \n- kidney 3 for validation \n\n###DataAugmentation\n- Flip (Horizontal and Vertical)\n- GAussian noise\n- gamma\n- scale rotate \n\n##Model\n-  tu-resnet50d UNET\n-  Effcientnet b0 UNET\n\nBest Performance \n- Public 0.805 Private 0.534 ensemble tu-resnet50D+ Efficientnet b0\n- Public 0.696 Private 0.610  tu-resnet50D\n\n"
  }
}