{
  "id": 457187,
  "title": "Pre-Print with lots of info on the training data and deep learning approaches to segmentation ",
  "url": "/competitions/blood-vessel-segmentation/discussion/457187",
  "author_name": "",
  "post_date": "2023-11-23T12:10:36.881684600Z",
  "votes": 38,
  "comment_count": 9,
  "views": 0,
  "content": "<p>Hi all, <br>\n   There have been many questions about the training data and about approaches to model choice. Here is a manuscript we have put together which reviews HiP-CT imaging and ML based vascular segmentation in the context of the current field. <br>\nIt also includes a benchmark model to see the type of baseline performance that can be achieved using just the training data you all have access to. It also highlights the area where the models most often fail. We put this together with the view to help and encourage you all with this competition, hopefully it achieves that. </p>\n<p><a href=\"https://doi.org/10.48550/arXiv.2311.13319\" target=\"_blank\">https://doi.org/10.48550/arXiv.2311.13319</a></p>\n<p>NOTE this paper does not include any training data that you all cannot access, and is tested on subsets of this same training data. i.e. NO public or private test data are in any way used in this manuscript. </p>",
  "messages": [
    {
      "id": "2535534",
      "postDate": "11/23/2023 12:10:36",
      "content": "<p>Hi all, <br>\n   There have been many questions about the training data and about approaches to model choice. Here is a manuscript we have put together which reviews HiP-CT imaging and ML based vascular segmentation in the context of the current field. <br>\nIt also includes a benchmark model to see the type of baseline performance that can be achieved using just the training data you all have access to. It also highlights the area where the models most often fail. We put this together with the view to help and encourage you all with this competition, hopefully it achieves that. </p>\n<p><a href=\"https://doi.org/10.48550/arXiv.2311.13319\" target=\"_blank\">https://doi.org/10.48550/arXiv.2311.13319</a></p>\n<p>NOTE this paper does not include any training data that you all cannot access, and is tested on subsets of this same training data. i.e. NO public or private test data are in any way used in this manuscript. </p>",
      "rawMarkdown": "Hi all, \n   There have been many questions about the training data and about approaches to model choice. Here is a manuscript we have put together which reviews HiP-CT imaging and ML based vascular segmentation in the context of the current field. \nIt also includes a benchmark model to see the type of baseline performance that can be achieved using just the training data you all have access to. It also highlights the area where the models most often fail. We put this together with the view to help and encourage you all with this competition, hopefully it achieves that. \n\nhttps://doi.org/10.48550/arXiv.2311.13319\n\nNOTE this paper does not include any training data that you all cannot access, and is tested on subsets of this same training data. i.e. NO public or private test data are in any way used in this manuscript.",
      "votes": null
    },
    {
      "id": "2540673",
      "postDate": "11/27/2023 21:42:45",
      "content": "<p>Thank you for sharing. This set a fairly ambitious baseline, should be a lot of fun trying to beat it :) </p>",
      "rawMarkdown": "Thank you for sharing. This set a fairly ambitious baseline, should be a lot of fun trying to beat it :)",
      "votes": null
    },
    {
      "id": "2543496",
      "postDate": "11/30/2023 06:15:24",
      "content": "<p>In the paper, the results of the experiments are based on dense or/and sparse annotations? </p>\n<p>e.g. we only have sparse annotations for kidney2.  </p>\n<p>Becuase different kidney 1,2,3 are annotated differently, so the results of the experiments needed to be interpreted with \"greater care\"?  </p>",
      "rawMarkdown": "In the paper, the results of the experiments are based on dense or/and sparse annotations? \n\ne.g. we only have sparse annotations for kidney2.  \n\nBecuase different kidney 1,2,3 are annotated differently, so the results of the experiments needed to be interpreted with \"greater care\"?",
      "votes": null
    },
    {
      "id": "2547892",
      "postDate": "12/04/2023 01:54:50",
      "content": "<p>does the testing set contain high-resolution subset (similar to <code>kidney_1_voi</code>)?</p>",
      "rawMarkdown": "does the testing set contain high-resolution subset (similar to `kidney_1_voi`)?",
      "votes": null
    },
    {
      "id": "2550514",
      "postDate": "12/06/2023 04:59:32",
      "content": "<p>Thank you, This has increased / upleveled the baseline</p>",
      "rawMarkdown": "Thank you, This has increased / upleveled the baseline",
      "votes": null
    },
    {
      "id": "2590866",
      "postDate": "01/07/2024 14:10:20",
      "content": "<p>I'm not quite familiar with the binning process, so I'd like to ask a question about it.😃</p>\n<p>When performing binning 2x, there will also be a resize operation on the mask. After the resize operation, the values will not be binary but rather between 0 and 1. How are these values, which are not quantized into binary, handled?</p>",
      "rawMarkdown": "I'm not quite familiar with the binning process, so I'd like to ask a question about it.😃\n\nWhen performing binning 2x, there will also be a resize operation on the mask. After the resize operation, the values will not be binary but rather between 0 and 1. How are these values, which are not quantized into binary, handled?",
      "votes": null
    },
    {
      "id": "2591009",
      "postDate": "01/07/2024 15:28:06",
      "content": "<p>Thank you for sharing. </p>",
      "rawMarkdown": "Thank you for sharing.",
      "votes": null
    },
    {
      "id": "2596204",
      "postDate": "01/10/2024 23:53:26",
      "content": "<p>Thanks for your sharing.</p>",
      "rawMarkdown": "Thanks for your sharing.",
      "votes": null
    },
    {
      "id": "2598990",
      "postDate": "01/12/2024 15:43:03",
      "content": "<p>Is this annotations good or bad?</p>",
      "rawMarkdown": "Is this annotations good or bad?",
      "votes": null
    },
    {
      "id": "2617404",
      "postDate": "01/24/2024 08:16:04",
      "content": "<p>Thank. This is very helpful.</p>",
      "rawMarkdown": "Thank. This is very helpful.",
      "votes": null
    }
  ],
  "comments": [
    {
      "id": 2540673,
      "author_name": "victorshlepov",
      "author_url": "",
      "post_date": "11/27/2023 21:42:45",
      "content": "<p>Thank you for sharing. This set a fairly ambitious baseline, should be a lot of fun trying to beat it :) </p>",
      "votes": null,
      "replies": []
    },
    {
      "id": 2543496,
      "author_name": "hengck23",
      "author_url": "",
      "post_date": "11/30/2023 06:15:24",
      "content": "<p>In the paper, the results of the experiments are based on dense or/and sparse annotations? </p>\n<p>e.g. we only have sparse annotations for kidney2.  </p>\n<p>Becuase different kidney 1,2,3 are annotated differently, so the results of the experiments needed to be interpreted with \"greater care\"?  </p>",
      "votes": null,
      "replies": [
        {
          "id": 2598990,
          "author_name": "zavodrobotov",
          "author_url": "",
          "post_date": "01/12/2024 15:43:03",
          "content": "<p>Is this annotations good or bad?</p>",
          "votes": null,
          "replies": []
        }
      ]
    },
    {
      "id": 2547892,
      "author_name": "maedward",
      "author_url": "",
      "post_date": "12/04/2023 01:54:50",
      "content": "<p>does the testing set contain high-resolution subset (similar to <code>kidney_1_voi</code>)?</p>",
      "votes": null,
      "replies": []
    },
    {
      "id": 2550514,
      "author_name": "ratnesh1729",
      "author_url": "",
      "post_date": "12/06/2023 04:59:32",
      "content": "<p>Thank you, This has increased / upleveled the baseline</p>",
      "votes": null,
      "replies": []
    },
    {
      "id": 2590866,
      "author_name": "siwooyong",
      "author_url": "",
      "post_date": "01/07/2024 14:10:20",
      "content": "<p>I'm not quite familiar with the binning process, so I'd like to ask a question about it.😃</p>\n<p>When performing binning 2x, there will also be a resize operation on the mask. After the resize operation, the values will not be binary but rather between 0 and 1. How are these values, which are not quantized into binary, handled?</p>",
      "votes": null,
      "replies": []
    },
    {
      "id": 2591009,
      "author_name": "kagglerdo",
      "author_url": "",
      "post_date": "01/07/2024 15:28:06",
      "content": "<p>Thank you for sharing. </p>",
      "votes": null,
      "replies": []
    },
    {
      "id": 2596204,
      "author_name": "mornicen",
      "author_url": "",
      "post_date": "01/10/2024 23:53:26",
      "content": "<p>Thanks for your sharing.</p>",
      "votes": null,
      "replies": []
    },
    {
      "id": 2617404,
      "author_name": "ronaldkhho",
      "author_url": "",
      "post_date": "01/24/2024 08:16:04",
      "content": "<p>Thank. This is very helpful.</p>",
      "votes": null,
      "replies": []
    }
  ],
  "raw_markdown_by_id": {
    "2535534": "Hi all, \n   There have been many questions about the training data and about approaches to model choice. Here is a manuscript we have put together which reviews HiP-CT imaging and ML based vascular segmentation in the context of the current field. \nIt also includes a benchmark model to see the type of baseline performance that can be achieved using just the training data you all have access to. It also highlights the area where the models most often fail. We put this together with the view to help and encourage you all with this competition, hopefully it achieves that. \n\nhttps://doi.org/10.48550/arXiv.2311.13319\n\nNOTE this paper does not include any training data that you all cannot access, and is tested on subsets of this same training data. i.e. NO public or private test data are in any way used in this manuscript.",
    "2540673": "Thank you for sharing. This set a fairly ambitious baseline, should be a lot of fun trying to beat it :)",
    "2543496": "In the paper, the results of the experiments are based on dense or/and sparse annotations? \n\ne.g. we only have sparse annotations for kidney2.  \n\nBecuase different kidney 1,2,3 are annotated differently, so the results of the experiments needed to be interpreted with \"greater care\"?",
    "2547892": "does the testing set contain high-resolution subset (similar to `kidney_1_voi`)?",
    "2550514": "Thank you, This has increased / upleveled the baseline",
    "2590866": "I'm not quite familiar with the binning process, so I'd like to ask a question about it.😃\n\nWhen performing binning 2x, there will also be a resize operation on the mask. After the resize operation, the values will not be binary but rather between 0 and 1. How are these values, which are not quantized into binary, handled?",
    "2591009": "Thank you for sharing.",
    "2596204": "Thanks for your sharing.",
    "2598990": "Is this annotations good or bad?",
    "2617404": "Thank. This is very helpful."
  },
  "source": "meta"
}