{
  "id": 234532,
  "title": "Bad specimen?",
  "url": "/competitions/herbarium-2021-fgvc8/discussion/234532",
  "author_name": "Kyle Oddson",
  "post_date": "2021-04-24T20:07:06.558000",
  "votes": 0,
  "comment_count": 2,
  "views": 0,
  "content": "<p>Image number 53500 in the test set doesn't appear to have a specimen in it…?</p>",
  "messages": [
    {
      "id": 1283977,
      "postDate": "2021-04-25T13:08:37.140Z",
      "content": "<p>Yeah, botanically speaking it seems like it's a herbarium specimen that only contains fragment packets. It's a proper specimen but comp. vision-wise I think it's probably impossible to get the right plant species given only these. </p>\n<p>I mean, humans/botanists wouldn't either if they can't see the labels nor look inside the packet 😄</p>",
      "rawMarkdown": "Yeah, botanically speaking it seems like it's a herbarium specimen that only contains fragment packets. It's a proper specimen but comp. vision-wise I think it's probably impossible to get the right plant species given only these. \n\nI mean, humans/botanists wouldn't either if they can't see the labels nor look inside the packet 😄",
      "votes": 1,
      "replies": [
        {
          "id": 1285757,
          "postDate": "2021-04-27T08:35:23.107Z",
          "content": "<p>Exactly! </p>\n<p>I will add that unfortunately it wasn't feasible to manually check all specimens in such a large dataset, further more for this competition we want to assess the \"real-world\" performance of computer vision models for herbarium datasets. </p>\n<p>This can also be an idea for a potential solution, to discard/down-weight certain specimens if they are non-representative.</p>\n<p>Good luck!!</p>",
          "rawMarkdown": "Exactly! \n\nI will add that unfortunately it wasn't feasible to manually check all specimens in such a large dataset, further more for this competition we want to assess the \"real-world\" performance of computer vision models for herbarium datasets. \n\nThis can also be an idea for a potential solution, to discard/down-weight certain specimens if they are non-representative.\n\nGood luck!!",
          "votes": 1
        }
      ]
    },
    {
      "id": 1283313,
      "postDate": "2021-04-24T20:07:06.560Z",
      "content": "<p>Image number 53500 in the test set doesn't appear to have a specimen in it…?</p>",
      "rawMarkdown": "Image number 53500 in the test set doesn't appear to have a specimen in it...?"
    }
  ],
  "comments": [
    {
      "id": 1283977,
      "author_name": "Dax Ledesma",
      "author_url": "",
      "post_date": "2021-04-25T13:08:37.140000",
      "content": "<p>Yeah, botanically speaking it seems like it's a herbarium specimen that only contains fragment packets. It's a proper specimen but comp. vision-wise I think it's probably impossible to get the right plant species given only these. </p>\n<p>I mean, humans/botanists wouldn't either if they can't see the labels nor look inside the packet 😄</p>",
      "votes": 1,
      "replies": [
        {
          "id": 1285757,
          "author_name": "Riccardo de Lutio",
          "author_url": "",
          "post_date": "2021-04-27T08:35:23.107000",
          "content": "<p>Exactly! </p>\n<p>I will add that unfortunately it wasn't feasible to manually check all specimens in such a large dataset, further more for this competition we want to assess the \"real-world\" performance of computer vision models for herbarium datasets. </p>\n<p>This can also be an idea for a potential solution, to discard/down-weight certain specimens if they are non-representative.</p>\n<p>Good luck!!</p>",
          "votes": 1,
          "replies": []
        }
      ]
    }
  ],
  "raw_markdown_by_id": {
    "1283977": "Yeah, botanically speaking it seems like it's a herbarium specimen that only contains fragment packets. It's a proper specimen but comp. vision-wise I think it's probably impossible to get the right plant species given only these. \n\nI mean, humans/botanists wouldn't either if they can't see the labels nor look inside the packet 😄",
    "1283313": "Image number 53500 in the test set doesn't appear to have a specimen in it...?"
  }
}