{
  "id": 614745,
  "title": "Important Info from a fellow Homo Sapien ",
  "url": "/competitions/csiro-biomass/discussion/614745",
  "author_name": "liyu zerihun",
  "post_date": "2025-11-06T07:03:28.278000",
  "votes": 3,
  "comment_count": 0,
  "views": 0,
  "content": "<p>This is just some extra info I want to share to help out newer contestants. </p>\n<p>Sample submission is not in the testing environment, if your code gets an exceptions that might be why. </p>\n<p>Tabular data is only used for training, not during testing. Meaning you can only use it when training your data, during testing they test using only the images. </p>\n<p>Your submission has to be a very specific format so be careful, follow the rules exactly(check data tab for more info). </p>\n<p>You can still use the tabular data, you just have to incorporate it into training. </p>\n<p>The actual number of training samples in the dataset is slightly more than 300(forgot the number). </p>\n<p>There is a huge imbalance between test and training, use split algorithms to get more accurate final val performance, so if you see your model performing above 0.70 with the vanilla approach it is most likely closer to 0.5. </p>",
  "messages": [
    {
      "id": 3311980,
      "postDate": "2025-11-06T07:03:28.277Z",
      "content": "<p>This is just some extra info I want to share to help out newer contestants. </p>\n<p>Sample submission is not in the testing environment, if your code gets an exceptions that might be why. </p>\n<p>Tabular data is only used for training, not during testing. Meaning you can only use it when training your data, during testing they test using only the images. </p>\n<p>Your submission has to be a very specific format so be careful, follow the rules exactly(check data tab for more info). </p>\n<p>You can still use the tabular data, you just have to incorporate it into training. </p>\n<p>The actual number of training samples in the dataset is slightly more than 300(forgot the number). </p>\n<p>There is a huge imbalance between test and training, use split algorithms to get more accurate final val performance, so if you see your model performing above 0.70 with the vanilla approach it is most likely closer to 0.5. </p>",
      "rawMarkdown": "This is just some extra info I want to share to help out newer contestants. \n\nSample submission is not in the testing environment, if your code gets an exceptions that might be why. \n\nTabular data is only used for training, not during testing. Meaning you can only use it when training your data, during testing they test using only the images. \n\nYour submission has to be a very specific format so be careful, follow the rules exactly(check data tab for more info). \n\nYou can still use the tabular data, you just have to incorporate it into training. \n\nThe actual number of training samples in the dataset is slightly more than 300(forgot the number). \n\nThere is a huge imbalance between test and training, use split algorithms to get more accurate final val performance, so if you see your model performing above 0.70 with the vanilla approach it is most likely closer to 0.5. \n\n\n\n\n",
      "votes": 3
    }
  ],
  "comments": [],
  "raw_markdown_by_id": {
    "3311980": "This is just some extra info I want to share to help out newer contestants. \n\nSample submission is not in the testing environment, if your code gets an exceptions that might be why. \n\nTabular data is only used for training, not during testing. Meaning you can only use it when training your data, during testing they test using only the images. \n\nYour submission has to be a very specific format so be careful, follow the rules exactly(check data tab for more info). \n\nYou can still use the tabular data, you just have to incorporate it into training. \n\nThe actual number of training samples in the dataset is slightly more than 300(forgot the number). \n\nThere is a huge imbalance between test and training, use split algorithms to get more accurate final val performance, so if you see your model performing above 0.70 with the vanilla approach it is most likely closer to 0.5. \n\n\n\n\n"
  }
}