{
  "id": 414582,
  "title": "8µm ground truth?",
  "url": "/competitions/vesuvius-challenge-ink-detection/discussion/414582",
  "author_name": "",
  "post_date": "2023-06-02T09:20:03.725690900Z",
  "votes": null,
  "comment_count": 1,
  "views": 0,
  "content": "<p>Hello!<br>\nI'm wondering is there any 8µm scans ground truth data?</p>\n<p>I tried to apply my model (trained on 4µm Kaggle challenge data) to some Scroll 2 segmented data. While I can read decent letters on Kaggle test samples, I read nothing on Scroll 2. Probably this is due to insufficient generalizing ability of my model, so it'd be logical to retrain it on \"Scroll 2\" data, for which I cannot found any ground truth.</p>\n<p>Thanks!</p>",
  "messages": [
    {
      "id": "2284793",
      "postDate": "06/02/2023 09:20:03",
      "content": "<p>Hello!<br>\nI'm wondering is there any 8µm scans ground truth data?</p>\n<p>I tried to apply my model (trained on 4µm Kaggle challenge data) to some Scroll 2 segmented data. While I can read decent letters on Kaggle test samples, I read nothing on Scroll 2. Probably this is due to insufficient generalizing ability of my model, so it'd be logical to retrain it on \"Scroll 2\" data, for which I cannot found any ground truth.</p>\n<p>Thanks!</p>",
      "rawMarkdown": "Hello!\nI'm wondering is there any 8µm scans ground truth data?\n\nI tried to apply my model (trained on 4µm Kaggle challenge data) to some Scroll 2 segmented data. While I can read decent letters on Kaggle test samples, I read nothing on Scroll 2. Probably this is due to insufficient generalizing ability of my model, so it'd be logical to retrain it on \"Scroll 2\" data, for which I cannot found any ground truth.\n\nThanks!",
      "votes": null
    },
    {
      "id": "2287388",
      "postDate": "06/04/2023 11:58:29",
      "content": "<p>And if you take every second for 4µm, will it not help?</p>",
      "rawMarkdown": "And if you take every second for 4µm, will it not help?",
      "votes": null
    }
  ],
  "comments": [
    {
      "id": 2287388,
      "author_name": "sapr3s",
      "author_url": "",
      "post_date": "06/04/2023 11:58:29",
      "content": "<p>And if you take every second for 4µm, will it not help?</p>",
      "votes": null,
      "replies": []
    }
  ],
  "raw_markdown_by_id": {
    "2284793": "Hello!\nI'm wondering is there any 8µm scans ground truth data?\n\nI tried to apply my model (trained on 4µm Kaggle challenge data) to some Scroll 2 segmented data. While I can read decent letters on Kaggle test samples, I read nothing on Scroll 2. Probably this is due to insufficient generalizing ability of my model, so it'd be logical to retrain it on \"Scroll 2\" data, for which I cannot found any ground truth.\n\nThanks!",
    "2287388": "And if you take every second for 4µm, will it not help?"
  },
  "source": "meta"
}