{
  "id": 64888,
  "title": "Run-Length Encoding -- rle_decode transposes results but rle_encode doesn't?",
  "url": "/competitions/airbus-ship-detection/discussion/64888",
  "author_name": "",
  "post_date": "2018-09-03T21:52:16.416051400Z",
  "votes": 1,
  "comment_count": 2,
  "views": 0,
  "content": "<p>I had a question about the earliest kernel that was provided by @inversion. He provided rle_encode and rle_decode functions. </p>\n\n<p>The decode function is doing a transpose operation at the very end with the comment: \"Needed to align to RLE direction\". It seems that the RLE of the training data is provided transposed by looking at the images and lining up the masks. </p>\n\n<p>However now the question i have is: should the encoded submission be also transposed? I haven't yet looked at the released test masks. Can anyone confirm whether or not rle_encode also needs to be transposed, or only when decoding the masks from the training set?</p>\n\n<p>Thanks!</p>",
  "messages": [
    {
      "id": "381002",
      "postDate": "09/03/2018 21:52:16",
      "content": "<p>I had a question about the earliest kernel that was provided by @inversion. He provided rle_encode and rle_decode functions. </p>\n\n<p>The decode function is doing a transpose operation at the very end with the comment: \"Needed to align to RLE direction\". It seems that the RLE of the training data is provided transposed by looking at the images and lining up the masks. </p>\n\n<p>However now the question i have is: should the encoded submission be also transposed? I haven't yet looked at the released test masks. Can anyone confirm whether or not rle_encode also needs to be transposed, or only when decoding the masks from the training set?</p>\n\n<p>Thanks!</p>",
      "rawMarkdown": "I had a question about the earliest kernel that was provided by @inversion. He provided rle_encode and rle_decode functions. \n\nThe decode function is doing a transpose operation at the very end with the comment: \"Needed to align to RLE direction\". It seems that the RLE of the training data is provided transposed by looking at the images and lining up the masks. \n\nHowever now the question i have is: should the encoded submission be also transposed? I haven't yet looked at the released test masks. Can anyone confirm whether or not rle_encode also needs to be transposed, or only when decoding the masks from the training set?\n\nThanks!",
      "votes": null
    },
    {
      "id": "381121",
      "postDate": "09/04/2018 05:01:01",
      "content": "<p>yes</p>",
      "rawMarkdown": "yes",
      "votes": null
    },
    {
      "id": "381435",
      "postDate": "09/04/2018 16:28:27",
      "content": "<p>Thanks! I'll try that next time I submit.</p>",
      "rawMarkdown": "Thanks! I'll try that next time I submit.",
      "votes": null
    }
  ],
  "comments": [
    {
      "id": 381121,
      "author_name": "alexiswoozcom",
      "author_url": "",
      "post_date": "09/04/2018 05:01:01",
      "content": "<p>yes</p>",
      "votes": null,
      "replies": [
        {
          "id": 381435,
          "author_name": "oewyn000",
          "author_url": "",
          "post_date": "09/04/2018 16:28:27",
          "content": "<p>Thanks! I'll try that next time I submit.</p>",
          "votes": null,
          "replies": []
        }
      ]
    }
  ],
  "raw_markdown_by_id": {
    "381002": "I had a question about the earliest kernel that was provided by @inversion. He provided rle_encode and rle_decode functions. \n\nThe decode function is doing a transpose operation at the very end with the comment: \"Needed to align to RLE direction\". It seems that the RLE of the training data is provided transposed by looking at the images and lining up the masks. \n\nHowever now the question i have is: should the encoded submission be also transposed? I haven't yet looked at the released test masks. Can anyone confirm whether or not rle_encode also needs to be transposed, or only when decoding the masks from the training set?\n\nThanks!",
    "381121": "yes",
    "381435": "Thanks! I'll try that next time I submit."
  },
  "source": "meta"
}