{
  "id": 70608,
  "title": "Offset between channels in dataset?",
  "url": "/competitions/human-protein-atlas-image-classification/discussion/70608",
  "author_name": "",
  "post_date": "2018-11-05T19:34:54.036862700Z",
  "votes": 3,
  "comment_count": 4,
  "views": 0,
  "content": "<p>I just accidentally found an image which has a strange offset between channels. </p>\n\n<p>I wonder how many more have this?</p>\n\n<p>The following images are from <code>6f062840-bba9-11e8-b2ba-ac1f6b6435d0</code>, ordered as R, G, B, Y:</p>\n\n<p><img src=\"https://pix.toile-libre.org/upload/original/1541446267.jpg\" alt=\"offset channels RGBY\"></p>\n\n<p>Has anyone checked for this?</p>\n\n<p><a href=\"https://pix.toile-libre.org/upload/original/1541446267.jpg\">https://pix.toile-libre.org/upload/original/1541446267.jpg</a></p>",
  "messages": [
    {
      "id": "415843",
      "postDate": "11/05/2018 19:34:54",
      "content": "<p>I just accidentally found an image which has a strange offset between channels. </p>\n\n<p>I wonder how many more have this?</p>\n\n<p>The following images are from <code>6f062840-bba9-11e8-b2ba-ac1f6b6435d0</code>, ordered as R, G, B, Y:</p>\n\n<p><img src=\"https://pix.toile-libre.org/upload/original/1541446267.jpg\" alt=\"offset channels RGBY\"></p>\n\n<p>Has anyone checked for this?</p>\n\n<p><a href=\"https://pix.toile-libre.org/upload/original/1541446267.jpg\">https://pix.toile-libre.org/upload/original/1541446267.jpg</a></p>",
      "rawMarkdown": "I just accidentally found an image which has a strange offset between channels. \n\nI wonder how many more have this?\n\nThe following images are from `6f062840-bba9-11e8-b2ba-ac1f6b6435d0`, ordered as R, G, B, Y:\n\n![offset channels RGBY](https://pix.toile-libre.org/upload/original/1541446267.jpg)\n\nHas anyone checked for this?\n\nhttps://pix.toile-libre.org/upload/original/1541446267.jpg",
      "votes": null
    },
    {
      "id": "415848",
      "postDate": "11/05/2018 19:58:31",
      "content": "<p>Someone bumped the microscope. Shouldn't be happening often, though likely more than once in 10s of thousands of images.</p>",
      "rawMarkdown": "Someone bumped the microscope. Shouldn't be happening often, though likely more than once in 10s of thousands of images.",
      "votes": null
    },
    {
      "id": "416703",
      "postDate": "11/07/2018 06:11:48",
      "content": "<p>Still, it'd be nice to check - perhaps its an error converting from tif. Not sure how to get statistics, though. Standard x correlation might not work. Maybe 31000 videos over color - take forever to QC. If it's 1/10000, how did he find it?</p>",
      "rawMarkdown": "Still, it'd be nice to check - perhaps its an error converting from tif. Not sure how to get statistics, though. Standard x correlation might not work. Maybe 31000 videos over color - take forever to QC. If it's 1/10000, how did he find it?",
      "votes": null
    },
    {
      "id": "417150",
      "postDate": "11/07/2018 20:47:37",
      "content": "<p>I looked at a few (50) images in an approximation to 4-frame videos(ubuntu has a nice image previewer built in) and didn't see any other shifts.\nI did see the shift that Daniel mentions. Looks like two shifts for that slide.\nI looked at the tiff version and it ALSO has the shift.</p>\n\n<p>Conclusion: the microscope bump theory looks good. But I have not proven that.</p>",
      "rawMarkdown": "I looked at a few (50) images in an approximation to 4-frame videos(ubuntu has a nice image previewer built in) and didn't see any other shifts.\nI did see the shift that Daniel mentions. Looks like two shifts for that slide.\nI looked at the tiff version and it ALSO has the shift.\n\nConclusion: the microscope bump theory looks good. But I have not proven that.",
      "votes": null
    },
    {
      "id": "417228",
      "postDate": "11/08/2018 00:55:39",
      "content": "<p>\"microscope bump theory\"</p>",
      "rawMarkdown": "\"microscope bump theory\"",
      "votes": null
    }
  ],
  "comments": [
    {
      "id": 415848,
      "author_name": "tilii7",
      "author_url": "",
      "post_date": "11/05/2018 19:58:31",
      "content": "<p>Someone bumped the microscope. Shouldn't be happening often, though likely more than once in 10s of thousands of images.</p>",
      "votes": null,
      "replies": []
    },
    {
      "id": 416703,
      "author_name": "petewills",
      "author_url": "",
      "post_date": "11/07/2018 06:11:48",
      "content": "<p>Still, it'd be nice to check - perhaps its an error converting from tif. Not sure how to get statistics, though. Standard x correlation might not work. Maybe 31000 videos over color - take forever to QC. If it's 1/10000, how did he find it?</p>",
      "votes": null,
      "replies": []
    },
    {
      "id": 417150,
      "author_name": "petewills",
      "author_url": "",
      "post_date": "11/07/2018 20:47:37",
      "content": "<p>I looked at a few (50) images in an approximation to 4-frame videos(ubuntu has a nice image previewer built in) and didn't see any other shifts.\nI did see the shift that Daniel mentions. Looks like two shifts for that slide.\nI looked at the tiff version and it ALSO has the shift.</p>\n\n<p>Conclusion: the microscope bump theory looks good. But I have not proven that.</p>",
      "votes": null,
      "replies": [
        {
          "id": 417228,
          "author_name": "kokecacao",
          "author_url": "",
          "post_date": "11/08/2018 00:55:39",
          "content": "<p>\"microscope bump theory\"</p>",
          "votes": null,
          "replies": []
        }
      ]
    }
  ],
  "raw_markdown_by_id": {
    "415843": "I just accidentally found an image which has a strange offset between channels. \n\nI wonder how many more have this?\n\nThe following images are from `6f062840-bba9-11e8-b2ba-ac1f6b6435d0`, ordered as R, G, B, Y:\n\n![offset channels RGBY](https://pix.toile-libre.org/upload/original/1541446267.jpg)\n\nHas anyone checked for this?\n\nhttps://pix.toile-libre.org/upload/original/1541446267.jpg",
    "415848": "Someone bumped the microscope. Shouldn't be happening often, though likely more than once in 10s of thousands of images.",
    "416703": "Still, it'd be nice to check - perhaps its an error converting from tif. Not sure how to get statistics, though. Standard x correlation might not work. Maybe 31000 videos over color - take forever to QC. If it's 1/10000, how did he find it?",
    "417150": "I looked at a few (50) images in an approximation to 4-frame videos(ubuntu has a nice image previewer built in) and didn't see any other shifts.\nI did see the shift that Daniel mentions. Looks like two shifts for that slide.\nI looked at the tiff version and it ALSO has the shift.\n\nConclusion: the microscope bump theory looks good. But I have not proven that.",
    "417228": "\"microscope bump theory\""
  },
  "source": "meta"
}