{
  "id": 505818,
  "title": "Camera Pose for the Transparent Category",
  "url": "/competitions/image-matching-challenge-2024/discussion/505818",
  "author_name": "",
  "post_date": "2024-05-19T08:34:26.152663800Z",
  "votes": 4,
  "comment_count": 1,
  "views": 0,
  "content": "<p>I am not 100% sure what the expected camera poses of the transparent category should be. The camera is obviously static with respect to the background, which means all groundtruth camera poses should be roughly the same -- this is aligned with the groundtruth camera poses provided. </p>\n<p>However, if this is the case, does it mean that if we encounter such cases during the test phase, we can just predict the camera to be static and can get the full score this way?</p>\n<p>And just to make sure, the camera pose will always be relative to the background (not the moving object in the image), right? </p>",
  "messages": [
    {
      "id": "2823501",
      "postDate": "05/19/2024 08:34:26",
      "content": "<p>I am not 100% sure what the expected camera poses of the transparent category should be. The camera is obviously static with respect to the background, which means all groundtruth camera poses should be roughly the same -- this is aligned with the groundtruth camera poses provided. </p>\n<p>However, if this is the case, does it mean that if we encounter such cases during the test phase, we can just predict the camera to be static and can get the full score this way?</p>\n<p>And just to make sure, the camera pose will always be relative to the background (not the moving object in the image), right? </p>",
      "rawMarkdown": "I am not 100% sure what the expected camera poses of the transparent category should be. The camera is obviously static with respect to the background, which means all groundtruth camera poses should be roughly the same -- this is aligned with the groundtruth camera poses provided. \n\nHowever, if this is the case, does it mean that if we encounter such cases during the test phase, we can just predict the camera to be static and can get the full score this way?\n\nAnd just to make sure, the camera pose will always be relative to the background (not the moving object in the image), right?",
      "votes": null
    },
    {
      "id": "2827190",
      "postDate": "05/21/2024 09:37:53",
      "content": "<p>If you visualize the training set camera poses, this may give you an idea. </p>",
      "rawMarkdown": "If you visualize the training set camera poses, this may give you an idea.",
      "votes": null
    }
  ],
  "comments": [
    {
      "id": 2827190,
      "author_name": "oldufo",
      "author_url": "",
      "post_date": "05/21/2024 09:37:53",
      "content": "<p>If you visualize the training set camera poses, this may give you an idea. </p>",
      "votes": null,
      "replies": []
    }
  ],
  "raw_markdown_by_id": {
    "2823501": "I am not 100% sure what the expected camera poses of the transparent category should be. The camera is obviously static with respect to the background, which means all groundtruth camera poses should be roughly the same -- this is aligned with the groundtruth camera poses provided. \n\nHowever, if this is the case, does it mean that if we encounter such cases during the test phase, we can just predict the camera to be static and can get the full score this way?\n\nAnd just to make sure, the camera pose will always be relative to the background (not the moving object in the image), right?",
    "2827190": "If you visualize the training set camera poses, this may give you an idea."
  },
  "source": "meta"
}