{
  "id": 36066,
  "title": "23 th place solution",
  "url": "/competitions/noaa-fisheries-steller-sea-lion-population-count/writeups/toshi-k-23-th-place-solution",
  "author_name": "",
  "post_date": "2017-11-13T10:46:21.907Z",
  "votes": 6,
  "comment_count": 2,
  "views": 0,
  "content": "<p>Hi Everyone,</p>\n\n<p>Ending final validation, I share my solution.<br> \nMy approach is based on semantic segmentation following by simple regression.<br>\nI remove full-connect layer from resnet 34, and use 7x7 buttom surface for segmentation.</p>\n\n<p>Code: <a href=\"https://github.com/toshi-k/kaggle-steller-sea-lion-population-count\">https://github.com/toshi-k/kaggle-steller-sea-lion-population-count</a></p>\n\n<p><img src=\"https://raw.githubusercontent.com/toshi-k/kaggle-steller-sea-lion-population-count/master/img/solution.png\" alt=\"enter image description here\" title=\"\"></p>",
  "messages": [
    {
      "id": "201110",
      "postDate": "07/10/2017 13:02:58",
      "content": "<p>Hi Everyone,</p>\n\n<p>Ending final validation, I share my solution.<br> \nMy approach is based on semantic segmentation following by simple regression.<br>\nI remove full-connect layer from resnet 34, and use 7x7 buttom surface for segmentation.</p>\n\n<p>Code: <a href=\"https://github.com/toshi-k/kaggle-steller-sea-lion-population-count\">https://github.com/toshi-k/kaggle-steller-sea-lion-population-count</a></p>\n\n<p><img src=\"https://raw.githubusercontent.com/toshi-k/kaggle-steller-sea-lion-population-count/master/img/solution.png\" alt=\"enter image description here\" title=\"\"></p>",
      "rawMarkdown": "Hi Everyone,\n\nEnding final validation, I share my solution.<br> \nMy approach is based on semantic segmentation following by simple regression.<br>\nI remove full-connect layer from resnet 34, and use 7x7 buttom surface for segmentation.\n\nCode: https://github.com/toshi-k/kaggle-steller-sea-lion-population-count\n\n![enter image description here][1]\n\n\n  [1]: https://raw.githubusercontent.com/toshi-k/kaggle-steller-sea-lion-population-count/master/img/solution.png",
      "votes": null
    },
    {
      "id": "210471",
      "postDate": "08/05/2017 21:28:24",
      "content": "<p>@Toshi_k,\nSorry for my late question:\nWould you explain what you called \"bias  sum(X) sum(X^2)\" in your model ? I looked at your code without success...\nThanks in advance for your answer.</p>",
      "rawMarkdown": "Toshi_k,\nSorry for my late question:\nWould you explain what you called \"bias  sum(X) sum(X^2)\" in your model ? I looked at your code without success...\nThanks in advance for your answer.",
      "votes": null
    },
    {
      "id": "210523",
      "postDate": "08/06/2017 03:39:17",
      "content": "<p>Hi eagle4,</p>\n\n<p>Result of semantic segmentation have [0~1] values for each pixel.<br>\nI summarize this result with three features and use for linear regression.<br>\nThis procedure is done in this code.<br>\n<a href=\"https://github.com/toshi-k/kaggle-steller-sea-lion-population-count/blob/master/source/02_deep/lib/extract.lua\">https://github.com/toshi-k/kaggle-steller-sea-lion-population-count/blob/master/source/02_deep/lib/extract.lua</a></p>",
      "rawMarkdown": "Hi eagle4,\n\nResult of semantic segmentation have [0~1] values for each pixel.<br>\nI summarize this result with three features and use for linear regression.<br>\nThis procedure is done in this code.<br>\nhttps://github.com/toshi-k/kaggle-steller-sea-lion-population-count/blob/master/source/02_deep/lib/extract.lua",
      "votes": null
    }
  ],
  "comments": [
    {
      "id": 210471,
      "author_name": "chabir",
      "author_url": "",
      "post_date": "08/05/2017 21:28:24",
      "content": "<p>@Toshi_k,\nSorry for my late question:\nWould you explain what you called \"bias  sum(X) sum(X^2)\" in your model ? I looked at your code without success...\nThanks in advance for your answer.</p>",
      "votes": null,
      "replies": []
    },
    {
      "id": 210523,
      "author_name": "toshik",
      "author_url": "",
      "post_date": "08/06/2017 03:39:17",
      "content": "<p>Hi eagle4,</p>\n\n<p>Result of semantic segmentation have [0~1] values for each pixel.<br>\nI summarize this result with three features and use for linear regression.<br>\nThis procedure is done in this code.<br>\n<a href=\"https://github.com/toshi-k/kaggle-steller-sea-lion-population-count/blob/master/source/02_deep/lib/extract.lua\">https://github.com/toshi-k/kaggle-steller-sea-lion-population-count/blob/master/source/02_deep/lib/extract.lua</a></p>",
      "votes": null,
      "replies": []
    }
  ],
  "raw_markdown_by_id": {
    "201110": "Hi Everyone,\n\nEnding final validation, I share my solution.<br> \nMy approach is based on semantic segmentation following by simple regression.<br>\nI remove full-connect layer from resnet 34, and use 7x7 buttom surface for segmentation.\n\nCode: https://github.com/toshi-k/kaggle-steller-sea-lion-population-count\n\n![enter image description here][1]\n\n\n  [1]: https://raw.githubusercontent.com/toshi-k/kaggle-steller-sea-lion-population-count/master/img/solution.png",
    "210471": "Toshi_k,\nSorry for my late question:\nWould you explain what you called \"bias  sum(X) sum(X^2)\" in your model ? I looked at your code without success...\nThanks in advance for your answer.",
    "210523": "Hi eagle4,\n\nResult of semantic segmentation have [0~1] values for each pixel.<br>\nI summarize this result with three features and use for linear regression.<br>\nThis procedure is done in this code.<br>\nhttps://github.com/toshi-k/kaggle-steller-sea-lion-population-count/blob/master/source/02_deep/lib/extract.lua"
  },
  "source": "meta"
}