{
  "id": 221071,
  "title": "one transformer idea",
  "url": "/competitions/ranzcr-clip-catheter-line-classification/discussion/221071",
  "author_name": "",
  "post_date": "2021-02-21T01:02:02.406295400Z",
  "votes": 8,
  "comment_count": 2,
  "views": 0,
  "content": "<p>this is just an idea</p>\n<p><img src=\"https://i.ibb.co/ZJ9W5R9/Selection-104.png\" alt=\"\"></p>\n<p>UP-DETR: Unsupervised Pre-training for Object Detection with Transformers<br>\n<a href=\"https://arxiv.org/abs/2011.09094\" target=\"_blank\">https://arxiv.org/abs/2011.09094</a><br>\n<img src=\"https://pbs.twimg.com/media/EnJv4n3WMAIfL8P.png\" alt=\"\"></p>\n<p><img src=\"https://i2.wp.com/syncedreview.com/wp-content/uploads/2020/11/image-98.png\" alt=\"\"></p>\n<p>related:<br>\nFew-Shot Object Detection with Attention-RPN and Multi-Relation Detector<br>\n<a href=\"https://arxiv.org/pdf/1908.01998v1.pdf\" target=\"_blank\">https://arxiv.org/pdf/1908.01998v1.pdf</a></p>\n<p><img src=\"https://d3i71xaburhd42.cloudfront.net/58d20519b190308d144d4b0110527231179ea100/1-Figure1-1.png\" alt=\"\"></p>\n<p>2 boxes at the top of image is the input query template</p>",
  "messages": [
    {
      "id": "1212179",
      "postDate": "02/21/2021 01:02:02",
      "content": "<p>this is just an idea</p>\n<p><img src=\"https://i.ibb.co/ZJ9W5R9/Selection-104.png\" alt=\"\"></p>\n<p>UP-DETR: Unsupervised Pre-training for Object Detection with Transformers<br>\n<a href=\"https://arxiv.org/abs/2011.09094\" target=\"_blank\">https://arxiv.org/abs/2011.09094</a><br>\n<img src=\"https://pbs.twimg.com/media/EnJv4n3WMAIfL8P.png\" alt=\"\"></p>\n<p><img src=\"https://i2.wp.com/syncedreview.com/wp-content/uploads/2020/11/image-98.png\" alt=\"\"></p>\n<p>related:<br>\nFew-Shot Object Detection with Attention-RPN and Multi-Relation Detector<br>\n<a href=\"https://arxiv.org/pdf/1908.01998v1.pdf\" target=\"_blank\">https://arxiv.org/pdf/1908.01998v1.pdf</a></p>\n<p><img src=\"https://d3i71xaburhd42.cloudfront.net/58d20519b190308d144d4b0110527231179ea100/1-Figure1-1.png\" alt=\"\"></p>\n<p>2 boxes at the top of image is the input query template</p>",
      "rawMarkdown": "this is just an idea\n\n![](https://i.ibb.co/ZJ9W5R9/Selection-104.png)\n\nUP-DETR: Unsupervised Pre-training for Object Detection with Transformers\nhttps://arxiv.org/abs/2011.09094\n![](https://pbs.twimg.com/media/EnJv4n3WMAIfL8P.png)\n\n![](https://i2.wp.com/syncedreview.com/wp-content/uploads/2020/11/image-98.png)\n\n\n\nrelated:\nFew-Shot Object Detection with Attention-RPN and Multi-Relation Detector\nhttps://arxiv.org/pdf/1908.01998v1.pdf\n\n![](https://d3i71xaburhd42.cloudfront.net/58d20519b190308d144d4b0110527231179ea100/1-Figure1-1.png)\n\n2 boxes at the top of image is the input query template",
      "votes": null
    },
    {
      "id": "1212191",
      "postDate": "02/21/2021 01:36:16",
      "content": "<p>Is a good idea but, any \"pretaining\" approach (on image data) is not suitable for someone who dosen't own a GPU, like me.</p>",
      "rawMarkdown": "Is a good idea but, any \"pretaining\" approach (on image data) is not suitable for someone who dosen't own a GPU, like me.",
      "votes": null
    },
    {
      "id": "1220043",
      "postDate": "02/27/2021 13:39:43",
      "content": "<p><a href=\"https://www.kaggle.com/hengck23\" target=\"_blank\">@hengck23</a> ,</p>\n<p>I was pursuing such a method for a while, thank you for sharing!<br>\nOne could think that there are so many combinations of pathologies and morphology-combinations that such a \"template matching\" is not  feasable… i think, that it would yield excellent results.</p>",
      "rawMarkdown": "hengck23 ,\n\nI was pursuing such a method for a while, thank you for sharing!\nOne could think that there are so many combinations of pathologies and morphology-combinations that such a \"template matching\" is not  feasable... i think, that it would yield excellent results.",
      "votes": null
    }
  ],
  "comments": [
    {
      "id": 1212191,
      "author_name": "hiramcho",
      "author_url": "",
      "post_date": "02/21/2021 01:36:16",
      "content": "<p>Is a good idea but, any \"pretaining\" approach (on image data) is not suitable for someone who dosen't own a GPU, like me.</p>",
      "votes": null,
      "replies": []
    },
    {
      "id": 1220043,
      "author_name": "sandorkonya",
      "author_url": "",
      "post_date": "02/27/2021 13:39:43",
      "content": "<p><a href=\"https://www.kaggle.com/hengck23\" target=\"_blank\">@hengck23</a> ,</p>\n<p>I was pursuing such a method for a while, thank you for sharing!<br>\nOne could think that there are so many combinations of pathologies and morphology-combinations that such a \"template matching\" is not  feasable… i think, that it would yield excellent results.</p>",
      "votes": null,
      "replies": []
    }
  ],
  "raw_markdown_by_id": {
    "1212179": "this is just an idea\n\n![](https://i.ibb.co/ZJ9W5R9/Selection-104.png)\n\nUP-DETR: Unsupervised Pre-training for Object Detection with Transformers\nhttps://arxiv.org/abs/2011.09094\n![](https://pbs.twimg.com/media/EnJv4n3WMAIfL8P.png)\n\n![](https://i2.wp.com/syncedreview.com/wp-content/uploads/2020/11/image-98.png)\n\n\n\nrelated:\nFew-Shot Object Detection with Attention-RPN and Multi-Relation Detector\nhttps://arxiv.org/pdf/1908.01998v1.pdf\n\n![](https://d3i71xaburhd42.cloudfront.net/58d20519b190308d144d4b0110527231179ea100/1-Figure1-1.png)\n\n2 boxes at the top of image is the input query template",
    "1212191": "Is a good idea but, any \"pretaining\" approach (on image data) is not suitable for someone who dosen't own a GPU, like me.",
    "1220043": "hengck23 ,\n\nI was pursuing such a method for a while, thank you for sharing!\nOne could think that there are so many combinations of pathologies and morphology-combinations that such a \"template matching\" is not  feasable... i think, that it would yield excellent results."
  },
  "source": "meta"
}