{
  "id": 308379,
  "title": "🤔  How to create a bounding boxes dataset ?",
  "url": "/competitions/happy-whale-and-dolphin/discussion/308379",
  "author_name": "Théo Boyer",
  "post_date": "2022-02-18T11:45:16.884000",
  "votes": 11,
  "comment_count": 0,
  "views": 0,
  "content": "<p>Hi kagglers 👋,<br>\nI'm convinced that a good cropping method will be required to have good results in this competition. I'm currently looking for ways to create a bounding boxes dataset for 🐳&amp;🐬.</p>\n<p>I've seen this amazing notebook: <a href=\"https://www.kaggle.com/awsaf49/happywhale-boundingbox-yolov5\" target=\"_blank\">Happywhale: BoundingBox [YOLOv5] 📦</a> by <a href=\"https://www.kaggle.com/awsaf49\" target=\"_blank\">@awsaf49</a> for a beginning of a solution.</p>\n<p>As this notebook leverages an old dataset that was crafted for the <a href=\"https://www.kaggle.com/c/whale-categorization-playground\" target=\"_blank\">whale tail identification compeition</a> there are still some major failure cases and numerous details that could be improved such as slight under/over-cropping present in a non-negligible part of the resulting dataset…</p>\n<p>So I'm sharing this notebook where I try to detect failure cases using OOD (Out Of Distribution) sample detection methods:<br>\n<a href=\"https://www.kaggle.com/wolfy73/filter-yolov5-failure-cases\" target=\"_blank\">🐳&amp;🐬 - Filter YOLOv5 failure cases</a></p>\n<p>It would be interesting to discuss new ways to obtain a high quality bounding boxes dataset for this competition, so do not hesitate to share your thoughts in the comments 👍 </p>",
  "messages": [
    {
      "id": 1695828,
      "postDate": "2022-02-18T11:45:16.883Z",
      "content": "<p>Hi kagglers 👋,<br>\nI'm convinced that a good cropping method will be required to have good results in this competition. I'm currently looking for ways to create a bounding boxes dataset for 🐳&amp;🐬.</p>\n<p>I've seen this amazing notebook: <a href=\"https://www.kaggle.com/awsaf49/happywhale-boundingbox-yolov5\" target=\"_blank\">Happywhale: BoundingBox [YOLOv5] 📦</a> by <a href=\"https://www.kaggle.com/awsaf49\" target=\"_blank\">@awsaf49</a> for a beginning of a solution.</p>\n<p>As this notebook leverages an old dataset that was crafted for the <a href=\"https://www.kaggle.com/c/whale-categorization-playground\" target=\"_blank\">whale tail identification compeition</a> there are still some major failure cases and numerous details that could be improved such as slight under/over-cropping present in a non-negligible part of the resulting dataset…</p>\n<p>So I'm sharing this notebook where I try to detect failure cases using OOD (Out Of Distribution) sample detection methods:<br>\n<a href=\"https://www.kaggle.com/wolfy73/filter-yolov5-failure-cases\" target=\"_blank\">🐳&amp;🐬 - Filter YOLOv5 failure cases</a></p>\n<p>It would be interesting to discuss new ways to obtain a high quality bounding boxes dataset for this competition, so do not hesitate to share your thoughts in the comments 👍 </p>",
      "rawMarkdown": "Hi kagglers 👋,\nI'm convinced that a good cropping method will be required to have good results in this competition. I'm currently looking for ways to create a bounding boxes dataset for 🐳&🐬.\n\nI've seen this amazing notebook: [Happywhale: BoundingBox [YOLOv5] 📦](https://www.kaggle.com/awsaf49/happywhale-boundingbox-yolov5) by @awsaf49 for a beginning of a solution.\n\nAs this notebook leverages an old dataset that was crafted for the [whale tail identification compeition](https://www.kaggle.com/c/whale-categorization-playground) there are still some major failure cases and numerous details that could be improved such as slight under/over-cropping present in a non-negligible part of the resulting dataset...\n\nSo I'm sharing this notebook where I try to detect failure cases using OOD (Out Of Distribution) sample detection methods:\n[🐳&🐬 - Filter YOLOv5 failure cases](https://www.kaggle.com/wolfy73/filter-yolov5-failure-cases)\n\nIt would be interesting to discuss new ways to obtain a high quality bounding boxes dataset for this competition, so do not hesitate to share your thoughts in the comments 👍 ",
      "votes": 11
    }
  ],
  "comments": [],
  "raw_markdown_by_id": {
    "1695828": "Hi kagglers 👋,\nI'm convinced that a good cropping method will be required to have good results in this competition. I'm currently looking for ways to create a bounding boxes dataset for 🐳&🐬.\n\nI've seen this amazing notebook: [Happywhale: BoundingBox [YOLOv5] 📦](https://www.kaggle.com/awsaf49/happywhale-boundingbox-yolov5) by @awsaf49 for a beginning of a solution.\n\nAs this notebook leverages an old dataset that was crafted for the [whale tail identification compeition](https://www.kaggle.com/c/whale-categorization-playground) there are still some major failure cases and numerous details that could be improved such as slight under/over-cropping present in a non-negligible part of the resulting dataset...\n\nSo I'm sharing this notebook where I try to detect failure cases using OOD (Out Of Distribution) sample detection methods:\n[🐳&🐬 - Filter YOLOv5 failure cases](https://www.kaggle.com/wolfy73/filter-yolov5-failure-cases)\n\nIt would be interesting to discuss new ways to obtain a high quality bounding boxes dataset for this competition, so do not hesitate to share your thoughts in the comments 👍 "
  }
}