{
  "id": 54665,
  "title": "Images in color but class labels are grayscale images ?",
  "url": "/competitions/cvpr-2018-autonomous-driving/discussion/54665",
  "author_name": "",
  "post_date": "2018-04-16T14:24:32.995601Z",
  "votes": null,
  "comment_count": 2,
  "views": 0,
  "content": "<p>Is it right that images are colour but class labels are grayscale images, \ndoenst make really sense, both pictures have to have RGB or Grayscale images.</p>\n\n<p>Best regards</p>",
  "messages": [
    {
      "id": "314931",
      "postDate": "04/16/2018 14:24:32",
      "content": "<p>Is it right that images are colour but class labels are grayscale images, \ndoenst make really sense, both pictures have to have RGB or Grayscale images.</p>\n\n<p>Best regards</p>",
      "rawMarkdown": "Is it right that images are colour but class labels are grayscale images, \ndoenst make really sense, both pictures have to have RGB or Grayscale images.\n\nBest regards",
      "votes": null
    },
    {
      "id": "315874",
      "postDate": "04/17/2018 17:57:32",
      "content": "<p>Hi Markus, i found(using opencv) that labels data are under 4 channel RGBa with a as Alpha Value. Unfortunately I cannot retrieve plain values that give me the pixel wise class number...</p>",
      "rawMarkdown": "Hi Markus, i found(using opencv) that labels data are under 4 channel RGBa with a as Alpha Value. Unfortunately I cannot retrieve plain values that give me the pixel wise class number...",
      "votes": null
    },
    {
      "id": "316105",
      "postDate": "04/18/2018 07:48:16",
      "content": "<p>Our input images are RGB color images. In the training set, we provide two kinds of grayscale output images, which are pixel-level annotations (8 bits) and instance-level annotations (16 bits). You may check the Data section for details.</p>",
      "rawMarkdown": "Our input images are RGB color images. In the training set, we provide two kinds of grayscale output images, which are pixel-level annotations (8 bits) and instance-level annotations (16 bits). You may check the Data section for details.",
      "votes": null
    }
  ],
  "comments": [
    {
      "id": 315874,
      "author_name": "dabriand",
      "author_url": "",
      "post_date": "04/17/2018 17:57:32",
      "content": "<p>Hi Markus, i found(using opencv) that labels data are under 4 channel RGBa with a as Alpha Value. Unfortunately I cannot retrieve plain values that give me the pixel wise class number...</p>",
      "votes": null,
      "replies": []
    },
    {
      "id": 316105,
      "author_name": "huangxinyu01",
      "author_url": "",
      "post_date": "04/18/2018 07:48:16",
      "content": "<p>Our input images are RGB color images. In the training set, we provide two kinds of grayscale output images, which are pixel-level annotations (8 bits) and instance-level annotations (16 bits). You may check the Data section for details.</p>",
      "votes": null,
      "replies": []
    }
  ],
  "raw_markdown_by_id": {
    "314931": "Is it right that images are colour but class labels are grayscale images, \ndoenst make really sense, both pictures have to have RGB or Grayscale images.\n\nBest regards",
    "315874": "Hi Markus, i found(using opencv) that labels data are under 4 channel RGBa with a as Alpha Value. Unfortunately I cannot retrieve plain values that give me the pixel wise class number...",
    "316105": "Our input images are RGB color images. In the training set, we provide two kinds of grayscale output images, which are pixel-level annotations (8 bits) and instance-level annotations (16 bits). You may check the Data section for details."
  },
  "source": "meta"
}