{
  "id": 404899,
  "title": "Question: translation scale",
  "url": "/competitions/image-matching-challenge-2023/discussion/404899",
  "author_name": "",
  "post_date": "2023-04-25T10:11:01.259501100Z",
  "votes": 3,
  "comment_count": 2,
  "views": 0,
  "content": "<p>As far as I know, evaluation metric is mean Average Accuracy (mAA) of every pair of images. <br>\nThen, when comparing the ground truth translation and estimated translation, how can we ensure that the scale is appropriately adjusted? <br>\nIs there a value related to the scale provided in the image set?</p>",
  "messages": [
    {
      "id": "2234582",
      "postDate": "04/25/2023 10:11:01",
      "content": "<p>As far as I know, evaluation metric is mean Average Accuracy (mAA) of every pair of images. <br>\nThen, when comparing the ground truth translation and estimated translation, how can we ensure that the scale is appropriately adjusted? <br>\nIs there a value related to the scale provided in the image set?</p>",
      "rawMarkdown": "As far as I know, evaluation metric is mean Average Accuracy (mAA) of every pair of images. \nThen, when comparing the ground truth translation and estimated translation, how can we ensure that the scale is appropriately adjusted? \nIs there a value related to the scale provided in the image set?",
      "votes": null
    },
    {
      "id": "2234606",
      "postDate": "04/25/2023 10:55:37",
      "content": "<p>We normalize the estimated translation to have unit-scale and then multiply it by the GT scale. You can find this in the metrics notebook.</p>",
      "rawMarkdown": "We normalize the estimated translation to have unit-scale and then multiply it by the GT scale. You can find this in the metrics notebook.",
      "votes": null
    },
    {
      "id": "2235386",
      "postDate": "04/26/2023 02:11:20",
      "content": "<p>Thank you so much!</p>",
      "rawMarkdown": "Thank you so much!",
      "votes": null
    }
  ],
  "comments": [
    {
      "id": 2234606,
      "author_name": "oldufo",
      "author_url": "",
      "post_date": "04/25/2023 10:55:37",
      "content": "<p>We normalize the estimated translation to have unit-scale and then multiply it by the GT scale. You can find this in the metrics notebook.</p>",
      "votes": null,
      "replies": [
        {
          "id": 2235386,
          "author_name": "wlsgur4011",
          "author_url": "",
          "post_date": "04/26/2023 02:11:20",
          "content": "<p>Thank you so much!</p>",
          "votes": null,
          "replies": []
        }
      ]
    }
  ],
  "raw_markdown_by_id": {
    "2234582": "As far as I know, evaluation metric is mean Average Accuracy (mAA) of every pair of images. \nThen, when comparing the ground truth translation and estimated translation, how can we ensure that the scale is appropriately adjusted? \nIs there a value related to the scale provided in the image set?",
    "2234606": "We normalize the estimated translation to have unit-scale and then multiply it by the GT scale. You can find this in the metrics notebook.",
    "2235386": "Thank you so much!"
  },
  "source": "meta"
}