{
  "id": 28414,
  "title": "Multi Category Classification at each pixel?",
  "url": "/competitions/dstl-satellite-imagery-feature-detection/discussion/28414",
  "author_name": "",
  "post_date": "2017-02-03T03:21:29.389667400Z",
  "votes": 2,
  "comment_count": 4,
  "views": 0,
  "content": "<p>Is this problem a multi category classification at each pixel?\nLike a single pixel could be both road and vehicle?</p>",
  "messages": [
    {
      "id": "159623",
      "postDate": "02/03/2017 03:21:29",
      "content": "<p>Is this problem a multi category classification at each pixel?\nLike a single pixel could be both road and vehicle?</p>",
      "rawMarkdown": "Is this problem a multi category classification at each pixel?\nLike a single pixel could be both road and vehicle?",
      "votes": null
    },
    {
      "id": "159685",
      "postDate": "02/03/2017 11:09:50",
      "content": "<p>Yes it is a multi classification problem, although most of the pixels are of one category</p>",
      "rawMarkdown": "Yes it is a multi classification problem, although most of the pixels are of one category",
      "votes": null
    },
    {
      "id": "159732",
      "postDate": "02/03/2017 16:53:04",
      "content": "<p>I have been treating this problem as a situation where overlaps are possible, but I have not tested for its existence.  It is exactly the thought of the car and the road that made me decide to assume overlaps might be possible. I was lazy though and did not check whether they do or do not overlap.</p>",
      "rawMarkdown": "I have been treating this problem as a situation where overlaps are possible, but I have not tested for its existence.  It is exactly the thought of the car and the road that made me decide to assume overlaps might be possible. I was lazy though and did not check whether they do or do not overlap.",
      "votes": null
    },
    {
      "id": "159734",
      "postDate": "02/03/2017 16:59:59",
      "content": "<p>Side note due to the way the submission file is set up, if you make the assumption that overlaps are possible you may never be any wiser( or I may never be any wiser.), because the submission file separates the classes to separate rows.</p>",
      "rawMarkdown": "Side note due to the way the submission file is set up, if you make the assumption that overlaps are possible you may never be any wiser( or I may never be any wiser.), because the submission file separates the classes to separate rows.",
      "votes": null
    },
    {
      "id": "159765",
      "postDate": "02/03/2017 19:26:23",
      "content": "<p>Yes, it is, multilabel classification, but you still can treat it as a multiclass and pick pixels that are above some threashold, say 0.4.</p>",
      "rawMarkdown": "Yes, it is, multilabel classification, but you still can treat it as a multiclass and pick pixels that are above some threashold, say 0.4.",
      "votes": null
    }
  ],
  "comments": [
    {
      "id": 159685,
      "author_name": "mxdbld",
      "author_url": "",
      "post_date": "02/03/2017 11:09:50",
      "content": "<p>Yes it is a multi classification problem, although most of the pixels are of one category</p>",
      "votes": null,
      "replies": []
    },
    {
      "id": 159732,
      "author_name": "joexjmmvhm",
      "author_url": "",
      "post_date": "02/03/2017 16:53:04",
      "content": "<p>I have been treating this problem as a situation where overlaps are possible, but I have not tested for its existence.  It is exactly the thought of the car and the road that made me decide to assume overlaps might be possible. I was lazy though and did not check whether they do or do not overlap.</p>",
      "votes": null,
      "replies": []
    },
    {
      "id": 159734,
      "author_name": "joexjmmvhm",
      "author_url": "",
      "post_date": "02/03/2017 16:59:59",
      "content": "<p>Side note due to the way the submission file is set up, if you make the assumption that overlaps are possible you may never be any wiser( or I may never be any wiser.), because the submission file separates the classes to separate rows.</p>",
      "votes": null,
      "replies": []
    },
    {
      "id": 159765,
      "author_name": "iglovikov",
      "author_url": "",
      "post_date": "02/03/2017 19:26:23",
      "content": "<p>Yes, it is, multilabel classification, but you still can treat it as a multiclass and pick pixels that are above some threashold, say 0.4.</p>",
      "votes": null,
      "replies": []
    }
  ],
  "raw_markdown_by_id": {
    "159623": "Is this problem a multi category classification at each pixel?\nLike a single pixel could be both road and vehicle?",
    "159685": "Yes it is a multi classification problem, although most of the pixels are of one category",
    "159732": "I have been treating this problem as a situation where overlaps are possible, but I have not tested for its existence.  It is exactly the thought of the car and the road that made me decide to assume overlaps might be possible. I was lazy though and did not check whether they do or do not overlap.",
    "159734": "Side note due to the way the submission file is set up, if you make the assumption that overlaps are possible you may never be any wiser( or I may never be any wiser.), because the submission file separates the classes to separate rows.",
    "159765": "Yes, it is, multilabel classification, but you still can treat it as a multiclass and pick pixels that are above some threashold, say 0.4."
  },
  "source": "meta"
}