{
  "id": 499167,
  "title": "2nd place solution",
  "url": "/competitions/spr-head-ct-age-prediction-challenge/discussion/499167",
  "author_name": "patriot",
  "post_date": "2024-04-30T23:25:35.802000",
  "votes": 4,
  "comment_count": 1,
  "views": 0,
  "content": "<p>preprocess: <a href=\"https://www.kaggle.com/code/abebe9849/sprheadct-preprocess\" target=\"_blank\">https://www.kaggle.com/code/abebe9849/sprheadct-preprocess</a></p>\n<p>train and inference:<a href=\"https://www.kaggle.com/code/abebe9849/sprheadct-train\" target=\"_blank\">https://www.kaggle.com/code/abebe9849/sprheadct-train</a></p>\n<p>slide: <a href=\"https://docs.google.com/presentation/d/1Y9hAUU0nrbgUWCjrYs_tHO06zPGP5Po5S788aRhHVIM/edit?usp=sharing\" target=\"_blank\">https://docs.google.com/presentation/d/1Y9hAUU0nrbgUWCjrYs_tHO06zPGP5Po5S788aRhHVIM/edit?usp=sharing</a><br>\n2.5dcnn like <a href=\"https://www.kaggle.com/competitions/prostate-cancer-grade-assessment/discussion/146855\" target=\"_blank\">PANDA</a>,backbone=efnetv2L<br>\nStack 3slice toRGB image<br>\n256×256×16~25slice <br>\nLoss=BCE(see <a href=\"https://www.kaggle.com/competitions/petfinder-pawpularity-score/discussion/275094\" target=\"_blank\">petfinder2</a>)<br>\n<a href=\"https://www.kaggle.com/code/shigemitsutomizawa/9th-place-swinb224-dldl-nn-svr-cat\" target=\"_blank\">DLDL2</a> loss not work for me (BCE&gt;CE=DLDL&gt;MAE)<br>\n10folds <a href=\"https://www.kaggle.com/competitions/ventilator-pressure-prediction/discussion/276138\" target=\"_blank\">median</a> ensemble</p>",
  "messages": [
    {
      "id": 2785601,
      "postDate": "2024-04-30T23:25:35.803Z",
      "content": "<p>preprocess: <a href=\"https://www.kaggle.com/code/abebe9849/sprheadct-preprocess\" target=\"_blank\">https://www.kaggle.com/code/abebe9849/sprheadct-preprocess</a></p>\n<p>train and inference:<a href=\"https://www.kaggle.com/code/abebe9849/sprheadct-train\" target=\"_blank\">https://www.kaggle.com/code/abebe9849/sprheadct-train</a></p>\n<p>slide: <a href=\"https://docs.google.com/presentation/d/1Y9hAUU0nrbgUWCjrYs_tHO06zPGP5Po5S788aRhHVIM/edit?usp=sharing\" target=\"_blank\">https://docs.google.com/presentation/d/1Y9hAUU0nrbgUWCjrYs_tHO06zPGP5Po5S788aRhHVIM/edit?usp=sharing</a><br>\n2.5dcnn like <a href=\"https://www.kaggle.com/competitions/prostate-cancer-grade-assessment/discussion/146855\" target=\"_blank\">PANDA</a>,backbone=efnetv2L<br>\nStack 3slice toRGB image<br>\n256×256×16~25slice <br>\nLoss=BCE(see <a href=\"https://www.kaggle.com/competitions/petfinder-pawpularity-score/discussion/275094\" target=\"_blank\">petfinder2</a>)<br>\n<a href=\"https://www.kaggle.com/code/shigemitsutomizawa/9th-place-swinb224-dldl-nn-svr-cat\" target=\"_blank\">DLDL2</a> loss not work for me (BCE&gt;CE=DLDL&gt;MAE)<br>\n10folds <a href=\"https://www.kaggle.com/competitions/ventilator-pressure-prediction/discussion/276138\" target=\"_blank\">median</a> ensemble</p>",
      "rawMarkdown": "preprocess: https://www.kaggle.com/code/abebe9849/sprheadct-preprocess\n\ntrain and inference:https://www.kaggle.com/code/abebe9849/sprheadct-train\n\nslide: https://docs.google.com/presentation/d/1Y9hAUU0nrbgUWCjrYs_tHO06zPGP5Po5S788aRhHVIM/edit?usp=sharing\n2.5dcnn like [PANDA](https://www.kaggle.com/competitions/prostate-cancer-grade-assessment/discussion/146855),backbone=efnetv2L\nStack 3slice toRGB image\n256×256×16~25slice \nLoss=BCE(see [petfinder2](https://www.kaggle.com/competitions/petfinder-pawpularity-score/discussion/275094))\n[DLDL2](https://www.kaggle.com/code/shigemitsutomizawa/9th-place-swinb224-dldl-nn-svr-cat) loss not work for me (BCE>CE=DLDL>MAE)\n10folds [median](https://www.kaggle.com/competitions/ventilator-pressure-prediction/discussion/276138) ensemble",
      "votes": 4
    },
    {
      "id": 2786197,
      "postDate": "2024-05-01T08:02:07.600Z",
      "content": "<pre><code>\n   \n   \n   \n   \n   \n \n\n   \n   \n \n\n   \n \n\n  \n     \n  \n     \n  \n     \n  \n     \n     \n     \n     \n  \n     \n     \n     \n  \n     \n  \n     \n  \n     \n  \n     \n     \n  \n     \n  \n     \n     \n     \n     \n\n   \n   \n   \n   \n   \n   \n\n   \n   \n   \n   \n   \n \n\n   \n   \n   \n   \n   \n   \n   \n   \n  \n     \n     \n     \n     \n</code></pre>\n<h2>private best config</h2>",
      "rawMarkdown": "```\ngeneral:\n  debug: false\n  exp_num: M004\n  device: 1\n  seed: 42\n  num_folds: 10\nN_patch: 25\nloss:\n  name: BCE\n  num_bins: 10\nprecrop: true\npreprocess:\n  size: 320\ngradient_checkpoint: false\naug:\n  HorizontalFlip:\n    p: 0.5\n  VerticalFlip:\n    p: 0.5\n  RandomRotate90:\n    p: 0.5\n  ShiftScaleRotate:\n    p: 0.7\n    shift_limit: 0.1\n    scale_limit: 0.1\n    rotate_limit: 15\n  RandomBrightnessContrast:\n    p: 0\n    brightness_limit: 0.1\n    contrast_limit: 0.1\n  CLAHE:\n    p: 0\n  one_of_Distortion:\n    p: 0\n  one_of_Blur_Gnoise:\n    p: 0\n  GridMask:\n    p: 0\n    num_grid: 2\n  compress:\n    p: 0\n  CoarseDropout:\n    p: 0.5\n    max_holes: 4\n    max_height: 6\n    max_width: 6\naugmentation:\n  do_mixup: false\n  do_fmix: false\n  do_cutmix: false\n  do_resizemix: false\n  mix_p: 0\n  mix_alpha: 1\nmodel:\n  name: tf_efficientnetv2_l\n  pooling: avg\n  drop_rate: 0.1\n  drop_path_rate: 0.1\n  stride: 2\npsuedo_label: 0\ntrain:\n  amp: true\n  amp_inf: false\n  optim: adamw\n  lr: 0.0001\n  epochs: 15\n  without_hesitate: 15\n  batch_size: 4\n  ga_accum: 1\n  scheduler:\n    name: cosine_warmup\n    min_lr: 1.0e-07\n    t_0: 3\n    warmup: 3\n```\n## private best config"
    }
  ],
  "comments": [
    {
      "id": 2786197,
      "author_name": "patriot",
      "author_url": "",
      "post_date": "2024-05-01T08:02:07.600000",
      "content": "<pre><code>\n   \n   \n   \n   \n   \n \n\n   \n   \n \n\n   \n \n\n  \n     \n  \n     \n  \n     \n  \n     \n     \n     \n     \n  \n     \n     \n     \n  \n     \n  \n     \n  \n     \n  \n     \n     \n  \n     \n  \n     \n     \n     \n     \n\n   \n   \n   \n   \n   \n   \n\n   \n   \n   \n   \n   \n \n\n   \n   \n   \n   \n   \n   \n   \n   \n  \n     \n     \n     \n     \n</code></pre>\n<h2>private best config</h2>",
      "votes": 0,
      "replies": []
    }
  ],
  "raw_markdown_by_id": {
    "2785601": "preprocess: https://www.kaggle.com/code/abebe9849/sprheadct-preprocess\n\ntrain and inference:https://www.kaggle.com/code/abebe9849/sprheadct-train\n\nslide: https://docs.google.com/presentation/d/1Y9hAUU0nrbgUWCjrYs_tHO06zPGP5Po5S788aRhHVIM/edit?usp=sharing\n2.5dcnn like [PANDA](https://www.kaggle.com/competitions/prostate-cancer-grade-assessment/discussion/146855),backbone=efnetv2L\nStack 3slice toRGB image\n256×256×16~25slice \nLoss=BCE(see [petfinder2](https://www.kaggle.com/competitions/petfinder-pawpularity-score/discussion/275094))\n[DLDL2](https://www.kaggle.com/code/shigemitsutomizawa/9th-place-swinb224-dldl-nn-svr-cat) loss not work for me (BCE>CE=DLDL>MAE)\n10folds [median](https://www.kaggle.com/competitions/ventilator-pressure-prediction/discussion/276138) ensemble",
    "2786197": "```\ngeneral:\n  debug: false\n  exp_num: M004\n  device: 1\n  seed: 42\n  num_folds: 10\nN_patch: 25\nloss:\n  name: BCE\n  num_bins: 10\nprecrop: true\npreprocess:\n  size: 320\ngradient_checkpoint: false\naug:\n  HorizontalFlip:\n    p: 0.5\n  VerticalFlip:\n    p: 0.5\n  RandomRotate90:\n    p: 0.5\n  ShiftScaleRotate:\n    p: 0.7\n    shift_limit: 0.1\n    scale_limit: 0.1\n    rotate_limit: 15\n  RandomBrightnessContrast:\n    p: 0\n    brightness_limit: 0.1\n    contrast_limit: 0.1\n  CLAHE:\n    p: 0\n  one_of_Distortion:\n    p: 0\n  one_of_Blur_Gnoise:\n    p: 0\n  GridMask:\n    p: 0\n    num_grid: 2\n  compress:\n    p: 0\n  CoarseDropout:\n    p: 0.5\n    max_holes: 4\n    max_height: 6\n    max_width: 6\naugmentation:\n  do_mixup: false\n  do_fmix: false\n  do_cutmix: false\n  do_resizemix: false\n  mix_p: 0\n  mix_alpha: 1\nmodel:\n  name: tf_efficientnetv2_l\n  pooling: avg\n  drop_rate: 0.1\n  drop_path_rate: 0.1\n  stride: 2\npsuedo_label: 0\ntrain:\n  amp: true\n  amp_inf: false\n  optim: adamw\n  lr: 0.0001\n  epochs: 15\n  without_hesitate: 15\n  batch_size: 4\n  ga_accum: 1\n  scheduler:\n    name: cosine_warmup\n    min_lr: 1.0e-07\n    t_0: 3\n    warmup: 3\n```\n## private best config"
  }
}