{
  "id": 297875,
  "title": "Concealed Object Detection",
  "url": "/competitions/tensorflow-great-barrier-reef/discussion/297875",
  "author_name": "",
  "post_date": "2021-12-30T09:18:34.446842200Z",
  "votes": 44,
  "comment_count": 1,
  "views": 0,
  "content": "<p>After a glance at the data, I feel the starfish is embedded well in the background and sometime it is hard to recognize even with my eyes. This remind me of Concealed Object Detection paper I read few months ago. Have anyone tried this approach?<br>\nPaper: <a href=\"https://arxiv.org/pdf/2102.10274.pdf\" target=\"_blank\">https://arxiv.org/pdf/2102.10274.pdf</a><br>\nCode: <a href=\"https://github.com/GewelsJI/SINet-V2\" target=\"_blank\">https://github.com/GewelsJI/SINet-V2</a><br>\n<img src=\"https://i.ibb.co/94FcPvW/concel.png\" alt=\"\"></p>",
  "messages": [
    {
      "id": "1632962",
      "postDate": "12/30/2021 09:18:34",
      "content": "<p>After a glance at the data, I feel the starfish is embedded well in the background and sometime it is hard to recognize even with my eyes. This remind me of Concealed Object Detection paper I read few months ago. Have anyone tried this approach?<br>\nPaper: <a href=\"https://arxiv.org/pdf/2102.10274.pdf\" target=\"_blank\">https://arxiv.org/pdf/2102.10274.pdf</a><br>\nCode: <a href=\"https://github.com/GewelsJI/SINet-V2\" target=\"_blank\">https://github.com/GewelsJI/SINet-V2</a><br>\n<img src=\"https://i.ibb.co/94FcPvW/concel.png\" alt=\"\"></p>",
      "rawMarkdown": "After a glance at the data, I feel the starfish is embedded well in the background and sometime it is hard to recognize even with my eyes. This remind me of Concealed Object Detection paper I read few months ago. Have anyone tried this approach?\nPaper: https://arxiv.org/pdf/2102.10274.pdf\nCode: https://github.com/GewelsJI/SINet-V2\n![](https://i.ibb.co/94FcPvW/concel.png)",
      "votes": null
    },
    {
      "id": "1636006",
      "postDate": "01/02/2022 13:55:45",
      "content": "<p>If I understood the article correctly, you need segmentation masks to train the model. But in our competition, unfortunately, there are only bounding boxes.</p>\n<p>But we can still try to use the Texture Enhanced Module from this paper. Thanks! </p>",
      "rawMarkdown": "If I understood the article correctly, you need segmentation masks to train the model. But in our competition, unfortunately, there are only bounding boxes.\n\nBut we can still try to use the Texture Enhanced Module from this paper. Thanks!",
      "votes": null
    }
  ],
  "comments": [
    {
      "id": 1636006,
      "author_name": "vadbeg",
      "author_url": "",
      "post_date": "01/02/2022 13:55:45",
      "content": "<p>If I understood the article correctly, you need segmentation masks to train the model. But in our competition, unfortunately, there are only bounding boxes.</p>\n<p>But we can still try to use the Texture Enhanced Module from this paper. Thanks! </p>",
      "votes": null,
      "replies": []
    }
  ],
  "raw_markdown_by_id": {
    "1632962": "After a glance at the data, I feel the starfish is embedded well in the background and sometime it is hard to recognize even with my eyes. This remind me of Concealed Object Detection paper I read few months ago. Have anyone tried this approach?\nPaper: https://arxiv.org/pdf/2102.10274.pdf\nCode: https://github.com/GewelsJI/SINet-V2\n![](https://i.ibb.co/94FcPvW/concel.png)",
    "1636006": "If I understood the article correctly, you need segmentation masks to train the model. But in our competition, unfortunately, there are only bounding boxes.\n\nBut we can still try to use the Texture Enhanced Module from this paper. Thanks!"
  },
  "source": "meta"
}