{
  "id": 681602,
  "title": "Supplementary Data: Pixel-level Semantic Segmentation Masks",
  "url": "/competitions/csiro-biomass/discussion/681602",
  "author_name": "Qiyu Liao",
  "post_date": "2026-03-16T06:51:08.397000",
  "votes": 2,
  "comment_count": 0,
  "views": 0,
  "content": "<p>I’m excited to share a supplementary dataset that provides <strong>pixel-level semantic annotations</strong> for the training set of the CSIRO - Image2Biomass Prediction competition. </p>\n<p>While the primary task is a regression challenge (predicting biomass weight), we believe that spatial information and pasture composition are key to improving model robustness. These masks are ideal for <strong>Multi-Task Learning (MTL)</strong> or for providing <strong>stronger auxiliary supervision</strong> during the feature extraction phase.</p>\n<p>You can find the annotation here:\n👉 <strong><a href=\"https://www.kaggle.com/datasets/qiyuliaocsiro/csiro-biomass-seg\" target=\"_blank\">CSIRO-Biomass-Seg Dataset</a></strong></p>",
  "messages": [
    {
      "id": 3421732,
      "postDate": "2026-03-16T06:51:08.397Z",
      "content": "<p>I’m excited to share a supplementary dataset that provides <strong>pixel-level semantic annotations</strong> for the training set of the CSIRO - Image2Biomass Prediction competition. </p>\n<p>While the primary task is a regression challenge (predicting biomass weight), we believe that spatial information and pasture composition are key to improving model robustness. These masks are ideal for <strong>Multi-Task Learning (MTL)</strong> or for providing <strong>stronger auxiliary supervision</strong> during the feature extraction phase.</p>\n<p>You can find the annotation here:\n👉 <strong><a href=\"https://www.kaggle.com/datasets/qiyuliaocsiro/csiro-biomass-seg\" target=\"_blank\">CSIRO-Biomass-Seg Dataset</a></strong></p>",
      "rawMarkdown": "I’m excited to share a supplementary dataset that provides **pixel-level semantic annotations** for the training set of the CSIRO - Image2Biomass Prediction competition. \n\nWhile the primary task is a regression challenge (predicting biomass weight), we believe that spatial information and pasture composition are key to improving model robustness. These masks are ideal for **Multi-Task Learning (MTL)** or for providing **stronger auxiliary supervision** during the feature extraction phase.\n\nYou can find the annotation here:\n👉 **[CSIRO-Biomass-Seg Dataset](https://www.kaggle.com/datasets/qiyuliaocsiro/csiro-biomass-seg)**",
      "votes": 2
    }
  ],
  "comments": [],
  "raw_markdown_by_id": {
    "3421732": "I’m excited to share a supplementary dataset that provides **pixel-level semantic annotations** for the training set of the CSIRO - Image2Biomass Prediction competition. \n\nWhile the primary task is a regression challenge (predicting biomass weight), we believe that spatial information and pasture composition are key to improving model robustness. These masks are ideal for **Multi-Task Learning (MTL)** or for providing **stronger auxiliary supervision** during the feature extraction phase.\n\nYou can find the annotation here:\n👉 **[CSIRO-Biomass-Seg Dataset](https://www.kaggle.com/datasets/qiyuliaocsiro/csiro-biomass-seg)**"
  }
}