{
  "id": 235590,
  "title": "Question about the bmp files in the test folder",
  "url": "/competitions/eurecom-aml-2021-challenge-2/discussion/235590",
  "author_name": "",
  "post_date": "2021-04-30T08:45:53.784529500Z",
  "votes": null,
  "comment_count": 2,
  "views": 0,
  "content": "<p>Hello, </p>\n<p>I understood during the meeting of Tuesday that the segmented files in the test folder are the result of the Baseline prediction? Or is it also the ground truth?</p>\n<p>Thanks for your help</p>",
  "messages": [
    {
      "id": "1288689",
      "postDate": "04/30/2021 08:45:53",
      "content": "<p>Hello, </p>\n<p>I understood during the meeting of Tuesday that the segmented files in the test folder are the result of the Baseline prediction? Or is it also the ground truth?</p>\n<p>Thanks for your help</p>",
      "rawMarkdown": "Hello, \n\nI understood during the meeting of Tuesday that the segmented files in the test folder are the result of the Baseline prediction? Or is it also the ground truth?\n\nThanks for your help",
      "votes": null
    },
    {
      "id": "1288734",
      "postDate": "04/30/2021 09:33:01",
      "content": "<p>Hi Hugo,</p>\n<p>Well this is a mistake on our part, these segmentations are ground truth segmentations and were not supposed to be given in the first place. </p>\n<p>Now, thinking about it, these ground truth segmentations could be useful to you if you want to evaluate your segmentation method on the test set and present results in the report. As we told you during the meeting this can be counted as bonus but should not be prioritised.</p>\n<p>So to sum up :</p>\n<ol>\n<li>You should NOT use the ground truth segmentations of the test set to perform any sort of training on it.</li>\n<li>You should NOT use the ground truth segmentations of the test set as input to your method in any way (computation of vCDR, localisation of a region of interest in the image, etc …).</li>\n<li>You CAN use these ground truth segmentations as a way to benchmark your segmentation method and expose the results in the report (for example to show that your superior results come from your better segmentation).</li>\n</ol>\n<p>Thank you for the remark !</p>\n<p>I am updating the rule section accordingly.</p>\n<p>Matthieu (TA)</p>",
      "rawMarkdown": "Hi Hugo,\n\nWell this is a mistake on our part, these segmentations are ground truth segmentations and were not supposed to be given in the first place. \n\nNow, thinking about it, these ground truth segmentations could be useful to you if you want to evaluate your segmentation method on the test set and present results in the report. As we told you during the meeting this can be counted as bonus but should not be prioritised.\n\nSo to sum up :\n1. You should NOT use the ground truth segmentations of the test set to perform any sort of training on it.\n2. You should NOT use the ground truth segmentations of the test set as input to your method in any way (computation of vCDR, localisation of a region of interest in the image, etc ...).\n3. You CAN use these ground truth segmentations as a way to benchmark your segmentation method and expose the results in the report (for example to show that your superior results come from your better segmentation).\n\nThank you for the remark !\n\nI am updating the rule section accordingly.\n\nMatthieu (TA)",
      "votes": null
    },
    {
      "id": "1288794",
      "postDate": "04/30/2021 10:44:31",
      "content": "<p>All clear thanks, yes it can be useful to understand where we can improve our work !</p>",
      "rawMarkdown": "All clear thanks, yes it can be useful to understand where we can improve our work !",
      "votes": null
    }
  ],
  "comments": [
    {
      "id": 1288734,
      "author_name": "mattdsf",
      "author_url": "",
      "post_date": "04/30/2021 09:33:01",
      "content": "<p>Hi Hugo,</p>\n<p>Well this is a mistake on our part, these segmentations are ground truth segmentations and were not supposed to be given in the first place. </p>\n<p>Now, thinking about it, these ground truth segmentations could be useful to you if you want to evaluate your segmentation method on the test set and present results in the report. As we told you during the meeting this can be counted as bonus but should not be prioritised.</p>\n<p>So to sum up :</p>\n<ol>\n<li>You should NOT use the ground truth segmentations of the test set to perform any sort of training on it.</li>\n<li>You should NOT use the ground truth segmentations of the test set as input to your method in any way (computation of vCDR, localisation of a region of interest in the image, etc …).</li>\n<li>You CAN use these ground truth segmentations as a way to benchmark your segmentation method and expose the results in the report (for example to show that your superior results come from your better segmentation).</li>\n</ol>\n<p>Thank you for the remark !</p>\n<p>I am updating the rule section accordingly.</p>\n<p>Matthieu (TA)</p>",
      "votes": null,
      "replies": [
        {
          "id": 1288794,
          "author_name": "cartesis07",
          "author_url": "",
          "post_date": "04/30/2021 10:44:31",
          "content": "<p>All clear thanks, yes it can be useful to understand where we can improve our work !</p>",
          "votes": null,
          "replies": []
        }
      ]
    }
  ],
  "raw_markdown_by_id": {
    "1288689": "Hello, \n\nI understood during the meeting of Tuesday that the segmented files in the test folder are the result of the Baseline prediction? Or is it also the ground truth?\n\nThanks for your help",
    "1288734": "Hi Hugo,\n\nWell this is a mistake on our part, these segmentations are ground truth segmentations and were not supposed to be given in the first place. \n\nNow, thinking about it, these ground truth segmentations could be useful to you if you want to evaluate your segmentation method on the test set and present results in the report. As we told you during the meeting this can be counted as bonus but should not be prioritised.\n\nSo to sum up :\n1. You should NOT use the ground truth segmentations of the test set to perform any sort of training on it.\n2. You should NOT use the ground truth segmentations of the test set as input to your method in any way (computation of vCDR, localisation of a region of interest in the image, etc ...).\n3. You CAN use these ground truth segmentations as a way to benchmark your segmentation method and expose the results in the report (for example to show that your superior results come from your better segmentation).\n\nThank you for the remark !\n\nI am updating the rule section accordingly.\n\nMatthieu (TA)",
    "1288794": "All clear thanks, yes it can be useful to understand where we can improve our work !"
  },
  "source": "meta"
}