{
  "id": 506874,
  "title": "Whats Working for you?",
  "url": "/competitions/image-matching-challenge-2024/discussion/506874",
  "author_name": "SHYAM GUPTA",
  "post_date": "2024-05-23T14:25:57.753000",
  "votes": 3,
  "comment_count": 1,
  "views": 0,
  "content": "<p>we have derived a lot of inspiration from IMC2023, but our score does not seem to increase above 0.13. Its my first vision competition so I am trying to learn as much as I can.</p>\n<p>we have tried, </p>\n<ol>\n<li>the standard pipeline from baseline notebooks and IMC2023 competitions. </li>\n<li>A global extractor with VIT backbone and  superglue with LOFTR for keypoints  matching<br>\nbut we are not able to get better.</li>\n</ol>\n<p>can someone tell what more I can do, to learn more about 3D scene reconstruction.</p>\n<p>are you using some kind of image augmentation? or some latest algorithms? what should we try to get better score in final submission. </p>",
  "messages": [
    {
      "id": 2831140,
      "postDate": "2024-05-23T14:25:57.753Z",
      "content": "<p>we have derived a lot of inspiration from IMC2023, but our score does not seem to increase above 0.13. Its my first vision competition so I am trying to learn as much as I can.</p>\n<p>we have tried, </p>\n<ol>\n<li>the standard pipeline from baseline notebooks and IMC2023 competitions. </li>\n<li>A global extractor with VIT backbone and  superglue with LOFTR for keypoints  matching<br>\nbut we are not able to get better.</li>\n</ol>\n<p>can someone tell what more I can do, to learn more about 3D scene reconstruction.</p>\n<p>are you using some kind of image augmentation? or some latest algorithms? what should we try to get better score in final submission. </p>",
      "rawMarkdown": "we have derived a lot of inspiration from IMC2023, but our score does not seem to increase above 0.13. Its my first vision competition so I am trying to learn as much as I can.\n\nwe have tried, \n1. the standard pipeline from baseline notebooks and IMC2023 competitions. \n2. A global extractor with VIT backbone and  superglue with LOFTR for keypoints  matching\nbut we are not able to get better.\n\ncan someone tell what more I can do, to learn more about 3D scene reconstruction.\n\nare you using some kind of image augmentation? or some latest algorithms? what should we try to get better score in final submission. \n",
      "votes": 3
    },
    {
      "id": 2840227,
      "postDate": "2024-05-28T03:18:05.047Z",
      "rawMarkdown": "",
      "isDeleted": true
    }
  ],
  "comments": [
    {
      "id": 2840227,
      "author_name": "",
      "author_url": "",
      "post_date": "2024-05-28T03:18:05.047000",
      "content": "",
      "votes": 0,
      "replies": []
    }
  ],
  "raw_markdown_by_id": {
    "2831140": "we have derived a lot of inspiration from IMC2023, but our score does not seem to increase above 0.13. Its my first vision competition so I am trying to learn as much as I can.\n\nwe have tried, \n1. the standard pipeline from baseline notebooks and IMC2023 competitions. \n2. A global extractor with VIT backbone and  superglue with LOFTR for keypoints  matching\nbut we are not able to get better.\n\ncan someone tell what more I can do, to learn more about 3D scene reconstruction.\n\nare you using some kind of image augmentation? or some latest algorithms? what should we try to get better score in final submission. \n",
    "2840227": ""
  }
}