{
  "id": 238049,
  "title": "46th Place Solution",
  "url": "/competitions/hubmap-kidney-segmentation/discussion/238049",
  "author_name": "makogarei",
  "post_date": "2021-05-11T04:08:30.317000",
  "votes": 3,
  "comment_count": 0,
  "views": 0,
  "content": "<h2>46th Place Solution</h2>\n<p>I would like to thank the organizers and management for organizing the competition.<br>\nIt was a fun way to deepen my knowledge.<br>\nAlso, if you are going to make this more of a competition, I would like to see a fuller description of the data generation process.<br>\nWe would also like to thank those who provided us with easy-to-understand kernels.</p>\n<hr>\n<h2>Pre processing:</h2>\n<ul>\n<li>resize tileimage size to (256, 256)</li>\n<li>pseudo label</li>\n<li>using d48_handlabel(Check the discussion.)</li>\n</ul>\n<h2>Augmentations:</h2>\n<p>I used albumentations.</p>\n<ul>\n<li>HorizontalFlip</li>\n<li>VerticalFlip</li>\n<li>ShiftScaleRotate, GridDistortion</li>\n<li>Blur, MedianBlur, GaussianBlur</li>\n<li>CLAHE, RandomBrightnessContrast, HueSaturationValue, IAASharpen</li>\n</ul>\n<h2>Post processing:</h2>\n<ul>\n<li>TTA(HorizontalFlip、VerticalFlip)</li>\n</ul>\n<h2>Model-1:</h2>\n<ul>\n<li>efficientnetb0 Unet</li>\n<li>image size : 256x256</li>\n<li>bce_dice（bce_weight：0.75）</li>\n<li>5FOLD</li>\n</ul>\n<h2>Model-2:</h2>\n<ul>\n<li>efficientnetb0 link</li>\n<li>image size : 256x256</li>\n<li>bce_dice（bce_weight：0.75）</li>\n<li>5FOLD</li>\n</ul>\n<h2>Model-3:</h2>\n<ul>\n<li>seresnext50 Unet</li>\n<li>image size : 256x256</li>\n<li>bce_dice（bce_weight：0.75）</li>\n<li>5FOLD</li>\n</ul>\n<h2>Model-4:</h2>\n<ul>\n<li>efficientnetb0 FPN</li>\n<li>image size : 256x256</li>\n<li>bce_dice（bce_weight：0.75）</li>\n<li>5FOLD</li>\n</ul>\n<h2>sub1:</h2>\n<ul>\n<li>1/4（Model-1+Model-2+Model-3+Model-4）</li>\n</ul>\n<p>I look forward to seeing you at another competition!</p>",
  "messages": [
    {
      "id": 1301398,
      "postDate": "2021-05-11T04:08:30.317Z",
      "content": "<h2>46th Place Solution</h2>\n<p>I would like to thank the organizers and management for organizing the competition.<br>\nIt was a fun way to deepen my knowledge.<br>\nAlso, if you are going to make this more of a competition, I would like to see a fuller description of the data generation process.<br>\nWe would also like to thank those who provided us with easy-to-understand kernels.</p>\n<hr>\n<h2>Pre processing:</h2>\n<ul>\n<li>resize tileimage size to (256, 256)</li>\n<li>pseudo label</li>\n<li>using d48_handlabel(Check the discussion.)</li>\n</ul>\n<h2>Augmentations:</h2>\n<p>I used albumentations.</p>\n<ul>\n<li>HorizontalFlip</li>\n<li>VerticalFlip</li>\n<li>ShiftScaleRotate, GridDistortion</li>\n<li>Blur, MedianBlur, GaussianBlur</li>\n<li>CLAHE, RandomBrightnessContrast, HueSaturationValue, IAASharpen</li>\n</ul>\n<h2>Post processing:</h2>\n<ul>\n<li>TTA(HorizontalFlip、VerticalFlip)</li>\n</ul>\n<h2>Model-1:</h2>\n<ul>\n<li>efficientnetb0 Unet</li>\n<li>image size : 256x256</li>\n<li>bce_dice（bce_weight：0.75）</li>\n<li>5FOLD</li>\n</ul>\n<h2>Model-2:</h2>\n<ul>\n<li>efficientnetb0 link</li>\n<li>image size : 256x256</li>\n<li>bce_dice（bce_weight：0.75）</li>\n<li>5FOLD</li>\n</ul>\n<h2>Model-3:</h2>\n<ul>\n<li>seresnext50 Unet</li>\n<li>image size : 256x256</li>\n<li>bce_dice（bce_weight：0.75）</li>\n<li>5FOLD</li>\n</ul>\n<h2>Model-4:</h2>\n<ul>\n<li>efficientnetb0 FPN</li>\n<li>image size : 256x256</li>\n<li>bce_dice（bce_weight：0.75）</li>\n<li>5FOLD</li>\n</ul>\n<h2>sub1:</h2>\n<ul>\n<li>1/4（Model-1+Model-2+Model-3+Model-4）</li>\n</ul>\n<p>I look forward to seeing you at another competition!</p>",
      "rawMarkdown": "## 46th Place Solution\nI would like to thank the organizers and management for organizing the competition.\nIt was a fun way to deepen my knowledge.\nAlso, if you are going to make this more of a competition, I would like to see a fuller description of the data generation process.\nWe would also like to thank those who provided us with easy-to-understand kernels.\n\n---\n\n## Pre processing:\n- resize tileimage size to (256, 256)\n- pseudo label\n- using d48_handlabel(Check the discussion.)\n\n## Augmentations:\nI used albumentations.\n- HorizontalFlip\n- VerticalFlip\n- ShiftScaleRotate, GridDistortion\n- Blur, MedianBlur, GaussianBlur\n- CLAHE, RandomBrightnessContrast, HueSaturationValue, IAASharpen\n\n## Post processing:\n- TTA(HorizontalFlip、VerticalFlip)\n\n## Model-1:\n- efficientnetb0 Unet\n- image size : 256x256\n- bce_dice（bce_weight：0.75）\n- 5FOLD\n\n## Model-2:\n- efficientnetb0 link\n- image size : 256x256\n- bce_dice（bce_weight：0.75）\n- 5FOLD\n\n## Model-3:\n- seresnext50 Unet\n- image size : 256x256\n- bce_dice（bce_weight：0.75）\n- 5FOLD\n\n## Model-4:\n- efficientnetb0 FPN\n- image size : 256x256\n- bce_dice（bce_weight：0.75）\n- 5FOLD\n\n## sub1:\n- 1/4（Model-1+Model-2+Model-3+Model-4）\n\nI look forward to seeing you at another competition!",
      "votes": 3
    }
  ],
  "comments": [],
  "raw_markdown_by_id": {
    "1301398": "## 46th Place Solution\nI would like to thank the organizers and management for organizing the competition.\nIt was a fun way to deepen my knowledge.\nAlso, if you are going to make this more of a competition, I would like to see a fuller description of the data generation process.\nWe would also like to thank those who provided us with easy-to-understand kernels.\n\n---\n\n## Pre processing:\n- resize tileimage size to (256, 256)\n- pseudo label\n- using d48_handlabel(Check the discussion.)\n\n## Augmentations:\nI used albumentations.\n- HorizontalFlip\n- VerticalFlip\n- ShiftScaleRotate, GridDistortion\n- Blur, MedianBlur, GaussianBlur\n- CLAHE, RandomBrightnessContrast, HueSaturationValue, IAASharpen\n\n## Post processing:\n- TTA(HorizontalFlip、VerticalFlip)\n\n## Model-1:\n- efficientnetb0 Unet\n- image size : 256x256\n- bce_dice（bce_weight：0.75）\n- 5FOLD\n\n## Model-2:\n- efficientnetb0 link\n- image size : 256x256\n- bce_dice（bce_weight：0.75）\n- 5FOLD\n\n## Model-3:\n- seresnext50 Unet\n- image size : 256x256\n- bce_dice（bce_weight：0.75）\n- 5FOLD\n\n## Model-4:\n- efficientnetb0 FPN\n- image size : 256x256\n- bce_dice（bce_weight：0.75）\n- 5FOLD\n\n## sub1:\n- 1/4（Model-1+Model-2+Model-3+Model-4）\n\nI look forward to seeing you at another competition!"
  }
}