{
  "id": 575483,
  "title": "Irregular samples",
  "url": "/competitions/beyond-visible-spectrum-ai-for-agriculture-2025/discussion/575483",
  "author_name": "",
  "post_date": "2025-04-28T23:34:51.867620800Z",
  "votes": 3,
  "comment_count": 1,
  "views": 0,
  "content": "<p>Hello, it appears that some of the samples in both training and testing sets have irregular shapes and others have a correct shape but complete missing data (total black channels). Is it an error in the dataset that should be therefor re-uploaded or is it something that we should handle ourselves? </p>",
  "messages": [
    {
      "id": "3189201",
      "postDate": "04/28/2025 23:34:51",
      "content": "<p>Hello, it appears that some of the samples in both training and testing sets have irregular shapes and others have a correct shape but complete missing data (total black channels). Is it an error in the dataset that should be therefor re-uploaded or is it something that we should handle ourselves? </p>",
      "rawMarkdown": "Hello, it appears that some of the samples in both training and testing sets have irregular shapes and others have a correct shape but complete missing data (total black channels). Is it an error in the dataset that should be therefor re-uploaded or is it something that we should handle ourselves?",
      "votes": null
    },
    {
      "id": "3189683",
      "postDate": "04/29/2025 15:38:30",
      "content": "<p>there are 100 (originally 125) channels per image. irregular shapes can be padded (immediate remedy) and one corrupt sample can be removed as it's only in the train set. </p>\n<p>synthetic data is mentioned as one of the tasks in the description, this probably decides the winning scores</p>",
      "rawMarkdown": "there are 100 (originally 125) channels per image. irregular shapes can be padded (immediate remedy) and one corrupt sample can be removed as it's only in the train set. \n\nsynthetic data is mentioned as one of the tasks in the description, this probably decides the winning scores",
      "votes": null
    }
  ],
  "comments": [
    {
      "id": 3189683,
      "author_name": "jaejohn",
      "author_url": "",
      "post_date": "04/29/2025 15:38:30",
      "content": "<p>there are 100 (originally 125) channels per image. irregular shapes can be padded (immediate remedy) and one corrupt sample can be removed as it's only in the train set. </p>\n<p>synthetic data is mentioned as one of the tasks in the description, this probably decides the winning scores</p>",
      "votes": null,
      "replies": []
    }
  ],
  "raw_markdown_by_id": {
    "3189201": "Hello, it appears that some of the samples in both training and testing sets have irregular shapes and others have a correct shape but complete missing data (total black channels). Is it an error in the dataset that should be therefor re-uploaded or is it something that we should handle ourselves?",
    "3189683": "there are 100 (originally 125) channels per image. irregular shapes can be padded (immediate remedy) and one corrupt sample can be removed as it's only in the train set. \n\nsynthetic data is mentioned as one of the tasks in the description, this probably decides the winning scores"
  },
  "source": "meta"
}