{
  "id": 574621,
  "title": "Zero Value Images in Test Set",
  "url": "/competitions/beyond-visible-spectrum-ai-for-agriculture-2025/discussion/574621",
  "author_name": "Wolf",
  "post_date": "2025-04-22T20:32:08.190000",
  "votes": 4,
  "comment_count": 4,
  "views": 0,
  "content": "<p>While trying to load the data from the ot folder i ran into quite a few samples with all zero elements. if this was just in the train set i would just drop them but since there are about 45 of them also in the test set i am not quite sure how to handle it or if this was even intended?</p>",
  "messages": [
    {
      "id": 3185107,
      "postDate": "2025-04-22T20:32:08.190Z",
      "content": "<p>While trying to load the data from the ot folder i ran into quite a few samples with all zero elements. if this was just in the train set i would just drop them but since there are about 45 of them also in the test set i am not quite sure how to handle it or if this was even intended?</p>",
      "rawMarkdown": "While trying to load the data from the ot folder i ran into quite a few samples with all zero elements. if this was just in the train set i would just drop them but since there are about 45 of them also in the test set i am not quite sure how to handle it or if this was even intended?",
      "votes": 4
    },
    {
      "id": 3193065,
      "postDate": "2025-05-03T18:59:00.217Z",
      "content": "<p>Hey! Sorry to bother, but I just saw your score on the competition and I’m honestly baffled-how did you get that result? I’ve tried everything, but my models do worse than just predicting the mean, and even basic data analysis seems to make things more random. Did you use a special trick or notice something in the data? Would really appreciate any insight!</p>",
      "rawMarkdown": "Hey! Sorry to bother, but I just saw your score on the competition and I’m honestly baffled-how did you get that result? I’ve tried everything, but my models do worse than just predicting the mean, and even basic data analysis seems to make things more random. Did you use a special trick or notice something in the data? Would really appreciate any insight!\n",
      "replies": [
        {
          "id": 3193310,
          "postDate": "2025-05-04T07:38:41.973Z",
          "rawMarkdown": "",
          "isDeleted": true
        },
        {
          "id": 3193318,
          "postDate": "2025-05-04T07:44:05.393Z",
          "content": "<p>hey! sorry to disappoint but so far i'm just padding the undersized images and then running everything resized to 224,224 through timm resnet models. it is really just the most basic approach. i also feel there is a problem with the data which made me resign from trying to figure anything out about it. </p>\n<p>edit: of course i'm playing around with some augmentations, freezing epochs, etc. but this is also just standard operations. </p>",
          "rawMarkdown": "hey! sorry to disappoint but so far i'm just padding the undersized images and then running everything resized to 224,224 through timm resnet models. it is really just the most basic approach. i also feel there is a problem with the data which made me resign from trying to figure anything out about it. \n\nedit: of course i'm playing around with some augmentations, freezing epochs, etc. but this is also just standard operations. "
        }
      ]
    },
    {
      "id": 3186875,
      "postDate": "2025-04-25T09:12:48.440Z",
      "content": "<p>Hi, it's normal that part of the patch 0. </p>",
      "rawMarkdown": "Hi, it's normal that part of the patch 0. "
    }
  ],
  "comments": [
    {
      "id": 3193065,
      "author_name": "Loric Bobon",
      "author_url": "",
      "post_date": "2025-05-03T18:59:00.217000",
      "content": "<p>Hey! Sorry to bother, but I just saw your score on the competition and I’m honestly baffled-how did you get that result? I’ve tried everything, but my models do worse than just predicting the mean, and even basic data analysis seems to make things more random. Did you use a special trick or notice something in the data? Would really appreciate any insight!</p>",
      "votes": 0,
      "replies": [
        {
          "id": 3193310,
          "author_name": "",
          "author_url": "",
          "post_date": "2025-05-04T07:38:41.973000",
          "content": "",
          "votes": 0,
          "replies": []
        },
        {
          "id": 3193318,
          "author_name": "Wolf",
          "author_url": "",
          "post_date": "2025-05-04T07:44:05.393000",
          "content": "<p>hey! sorry to disappoint but so far i'm just padding the undersized images and then running everything resized to 224,224 through timm resnet models. it is really just the most basic approach. i also feel there is a problem with the data which made me resign from trying to figure anything out about it. </p>\n<p>edit: of course i'm playing around with some augmentations, freezing epochs, etc. but this is also just standard operations. </p>",
          "votes": 0,
          "replies": []
        }
      ]
    },
    {
      "id": 3186875,
      "author_name": "robeson",
      "author_url": "",
      "post_date": "2025-04-25T09:12:48.440000",
      "content": "<p>Hi, it's normal that part of the patch 0. </p>",
      "votes": 0,
      "replies": []
    }
  ],
  "raw_markdown_by_id": {
    "3185107": "While trying to load the data from the ot folder i ran into quite a few samples with all zero elements. if this was just in the train set i would just drop them but since there are about 45 of them also in the test set i am not quite sure how to handle it or if this was even intended?",
    "3193065": "Hey! Sorry to bother, but I just saw your score on the competition and I’m honestly baffled-how did you get that result? I’ve tried everything, but my models do worse than just predicting the mean, and even basic data analysis seems to make things more random. Did you use a special trick or notice something in the data? Would really appreciate any insight!\n",
    "3186875": "Hi, it's normal that part of the patch 0. "
  }
}