{
  "id": 117268,
  "title": "Position3 postprocessing",
  "url": "/competitions/rsna-intracranial-hemorrhage-detection/discussion/117268",
  "author_name": "Aimoldin Anuar [dsmlkz]",
  "post_date": "2019-11-14T08:32:56.606000",
  "votes": 3,
  "comment_count": 0,
  "views": 0,
  "content": "<p>Write <strong>kernel</strong> about our postprocessing technics:\n<a href=\"https://www.kaggle.com/sneddy/postprocessing-approach\">https://www.kaggle.com/sneddy/postprocessing-approach</a></p>\n\n<p>You can try on your best submit  </p>\n\n<p><strong>Idea:</strong> sort by Position3 and apply fixed 1x3 filters for each channel.\nFor this, for each image, its neighbor was calculated on the left and on the right. Then these filters were applied several times, seriously smoothing the result.</p>\n\n<p><strong>Dataset</strong> with precalculated neigbours provided here:\n<a href=\"https://www.kaggle.com/sneddy/hemoorrhagemetadata\">https://www.kaggle.com/sneddy/hemoorrhagemetadata</a></p>\n\n<p>This technology improved the 1stage-leaderboard competition metric by a value in the range of 0.002 - 0.01</p>\n\n<p>The following is an example of applying this filter to the test data from the first and second stages.</p>",
  "messages": [
    {
      "id": 672863,
      "postDate": "2019-11-14T08:32:56.607Z",
      "content": "<p>Write <strong>kernel</strong> about our postprocessing technics:\n<a href=\"https://www.kaggle.com/sneddy/postprocessing-approach\">https://www.kaggle.com/sneddy/postprocessing-approach</a></p>\n\n<p>You can try on your best submit  </p>\n\n<p><strong>Idea:</strong> sort by Position3 and apply fixed 1x3 filters for each channel.\nFor this, for each image, its neighbor was calculated on the left and on the right. Then these filters were applied several times, seriously smoothing the result.</p>\n\n<p><strong>Dataset</strong> with precalculated neigbours provided here:\n<a href=\"https://www.kaggle.com/sneddy/hemoorrhagemetadata\">https://www.kaggle.com/sneddy/hemoorrhagemetadata</a></p>\n\n<p>This technology improved the 1stage-leaderboard competition metric by a value in the range of 0.002 - 0.01</p>\n\n<p>The following is an example of applying this filter to the test data from the first and second stages.</p>",
      "rawMarkdown": "Write **kernel** about our postprocessing technics:\nhttps://www.kaggle.com/sneddy/postprocessing-approach\n\nYou can try on your best submit  \n\n**Idea:** sort by Position3 and apply fixed 1x3 filters for each channel.\nFor this, for each image, its neighbor was calculated on the left and on the right. Then these filters were applied several times, seriously smoothing the result.\n\n**Dataset** with precalculated neigbours provided here:\nhttps://www.kaggle.com/sneddy/hemoorrhagemetadata\n\nThis technology improved the 1stage-leaderboard competition metric by a value in the range of 0.002 - 0.01\n\nThe following is an example of applying this filter to the test data from the first and second stages.",
      "votes": 3
    }
  ],
  "comments": [],
  "raw_markdown_by_id": {
    "672863": "Write **kernel** about our postprocessing technics:\nhttps://www.kaggle.com/sneddy/postprocessing-approach\n\nYou can try on your best submit  \n\n**Idea:** sort by Position3 and apply fixed 1x3 filters for each channel.\nFor this, for each image, its neighbor was calculated on the left and on the right. Then these filters were applied several times, seriously smoothing the result.\n\n**Dataset** with precalculated neigbours provided here:\nhttps://www.kaggle.com/sneddy/hemoorrhagemetadata\n\nThis technology improved the 1stage-leaderboard competition metric by a value in the range of 0.002 - 0.01\n\nThe following is an example of applying this filter to the test data from the first and second stages."
  }
}