{
  "id": 68454,
  "title": "Missing Images of target ID 10, 15 and 17",
  "url": "/competitions/human-protein-atlas-image-classification/discussion/68454",
  "author_name": "",
  "post_date": "2018-10-12T23:10:59.932452100Z",
  "votes": 1,
  "comment_count": 6,
  "views": 0,
  "content": "<p>Hello everyone,</p>\n\n<p>I have noticed that there's no images of target ID 10, 15 and 17. In table.png you can see the resume of how many images are for each target ID</p>\n\n<p>Would it possible that Human Protein Atlas uploads images of these targets ID?\nAlso adding images for targets IDs: 9 and 27 will be great for a better accuracy.</p>\n\n<p>Best Regards,\nJulian Vidal</p>",
  "messages": [
    {
      "id": "403117",
      "postDate": "10/12/2018 23:10:59",
      "content": "<p>Hello everyone,</p>\n\n<p>I have noticed that there's no images of target ID 10, 15 and 17. In table.png you can see the resume of how many images are for each target ID</p>\n\n<p>Would it possible that Human Protein Atlas uploads images of these targets ID?\nAlso adding images for targets IDs: 9 and 27 will be great for a better accuracy.</p>\n\n<p>Best Regards,\nJulian Vidal</p>",
      "rawMarkdown": "Hello everyone,\n\nI have noticed that there's no images of target ID 10, 15 and 17. In table.png you can see the resume of how many images are for each target ID\n\nWould it possible that Human Protein Atlas uploads images of these targets ID?\nAlso adding images for targets IDs: 9 and 27 will be great for a better accuracy.\n\nBest Regards,\nJulian Vidal",
      "votes": null
    },
    {
      "id": "403137",
      "postDate": "10/13/2018 00:34:26",
      "content": "<p>Probably, you read only first class in list of classes, because all images are in, e.g. image 01314a58-bba3-11e8-b2b9-ac1f6b6435d0 has 10th class. Also, you can see correct distribution in others kernels (e.g. <a href=\"https://www.kaggle.com/allunia/protein-atlas-exporation-and-baseline\">https://www.kaggle.com/allunia/protein-atlas-exporation-and-baseline</a>).</p>",
      "rawMarkdown": "Probably, you read only first class in list of classes, because all images are in, e.g. image 01314a58-bba3-11e8-b2b9-ac1f6b6435d0 has 10th class. Also, you can see correct distribution in others kernels (e.g. https://www.kaggle.com/allunia/protein-atlas-exporation-and-baseline).",
      "votes": null
    },
    {
      "id": "404314",
      "postDate": "10/15/2018 14:56:28",
      "content": "<p>I checked and I have no empty class. </p>\n\n<p>Note that some images have 2 or 3 classes (maybe more)</p>\n\n<p>Yes, I agree we should have more images for certain classes, especially 27, which is really really lacking. </p>",
      "rawMarkdown": "I checked and I have no empty class. \n\nNote that some images have 2 or 3 classes (maybe more)\n\nYes, I agree we should have more images for certain classes, especially 27, which is really really lacking.",
      "votes": null
    },
    {
      "id": "405060",
      "postDate": "10/16/2018 19:46:43",
      "content": "<p>I mean that there are no images that ONLY have 1 target ID 10, 15 and 17. I mean, e.g 5dbfaf28-bba6-11e8-b2ba-ac1f6b6435d0 has target ID 10 and 9. but there is no any image with ONLY ID target 10</p>",
      "rawMarkdown": "I mean that there are no images that ONLY have 1 target ID 10, 15 and 17. I mean, e.g 5dbfaf28-bba6-11e8-b2ba-ac1f6b6435d0 has target ID 10 and 9. but there is no any image with ONLY ID target 10",
      "votes": null
    },
    {
      "id": "405061",
      "postDate": "10/16/2018 19:48:03",
      "content": "<p>Can you please give the name of an image that ONLY has ID target 10, 15 or 17? Thanks a lot!</p>",
      "rawMarkdown": "Can you please give the name of an image that ONLY has ID target 10, 15 or 17? Thanks a lot!",
      "votes": null
    },
    {
      "id": "405160",
      "postDate": "10/17/2018 01:27:14",
      "content": "<p>I didn't calculate that, but I'm sure there are images with those IDs, and they're more than images with ID 27.</p>\n\n<p>If we are supposing it right about your table (you got the first ID instead of all IDs), there isn't an image with \"only\" one of those classes. </p>",
      "rawMarkdown": "I didn't calculate that, but I'm sure there are images with those IDs, and they're more than images with ID 27.\n\nIf we are supposing it right about your table (you got the first ID instead of all IDs), there isn't an image with \"only\" one of those classes.",
      "votes": null
    },
    {
      "id": "407952",
      "postDate": "10/22/2018 03:47:23",
      "content": "<p>It is now proven that you could not use this as a rule to \"post-process\" results. As <a href=\"https://www.kaggle.com/c/human-protein-atlas-image-classification/discussion/68678\">LB probing</a> points out, there are classes in LB only have 17. But 10, 15 cannot be certain because of the round down error. Hope it helps. :)</p>",
      "rawMarkdown": "It is now proven that you could not use this as a rule to \"post-process\" results. As [LB probing][1] points out, there are classes in LB only have 17. But 10, 15 cannot be certain because of the round down error. Hope it helps. :)\n\n\n  [1]: https://www.kaggle.com/c/human-protein-atlas-image-classification/discussion/68678",
      "votes": null
    }
  ],
  "comments": [
    {
      "id": 403137,
      "author_name": "kwentar",
      "author_url": "",
      "post_date": "10/13/2018 00:34:26",
      "content": "<p>Probably, you read only first class in list of classes, because all images are in, e.g. image 01314a58-bba3-11e8-b2b9-ac1f6b6435d0 has 10th class. Also, you can see correct distribution in others kernels (e.g. <a href=\"https://www.kaggle.com/allunia/protein-atlas-exporation-and-baseline\">https://www.kaggle.com/allunia/protein-atlas-exporation-and-baseline</a>).</p>",
      "votes": null,
      "replies": [
        {
          "id": 405060,
          "author_name": "ijulian97",
          "author_url": "",
          "post_date": "10/16/2018 19:46:43",
          "content": "<p>I mean that there are no images that ONLY have 1 target ID 10, 15 and 17. I mean, e.g 5dbfaf28-bba6-11e8-b2ba-ac1f6b6435d0 has target ID 10 and 9. but there is no any image with ONLY ID target 10</p>",
          "votes": null,
          "replies": []
        }
      ]
    },
    {
      "id": 404314,
      "author_name": "danmoller",
      "author_url": "",
      "post_date": "10/15/2018 14:56:28",
      "content": "<p>I checked and I have no empty class. </p>\n\n<p>Note that some images have 2 or 3 classes (maybe more)</p>\n\n<p>Yes, I agree we should have more images for certain classes, especially 27, which is really really lacking. </p>",
      "votes": null,
      "replies": [
        {
          "id": 405061,
          "author_name": "ijulian97",
          "author_url": "",
          "post_date": "10/16/2018 19:48:03",
          "content": "<p>Can you please give the name of an image that ONLY has ID target 10, 15 or 17? Thanks a lot!</p>",
          "votes": null,
          "replies": []
        },
        {
          "id": 405160,
          "author_name": "danmoller",
          "author_url": "",
          "post_date": "10/17/2018 01:27:14",
          "content": "<p>I didn't calculate that, but I'm sure there are images with those IDs, and they're more than images with ID 27.</p>\n\n<p>If we are supposing it right about your table (you got the first ID instead of all IDs), there isn't an image with \"only\" one of those classes. </p>",
          "votes": null,
          "replies": []
        }
      ]
    },
    {
      "id": 407952,
      "author_name": "kokecacao",
      "author_url": "",
      "post_date": "10/22/2018 03:47:23",
      "content": "<p>It is now proven that you could not use this as a rule to \"post-process\" results. As <a href=\"https://www.kaggle.com/c/human-protein-atlas-image-classification/discussion/68678\">LB probing</a> points out, there are classes in LB only have 17. But 10, 15 cannot be certain because of the round down error. Hope it helps. :)</p>",
      "votes": null,
      "replies": []
    }
  ],
  "raw_markdown_by_id": {
    "403117": "Hello everyone,\n\nI have noticed that there's no images of target ID 10, 15 and 17. In table.png you can see the resume of how many images are for each target ID\n\nWould it possible that Human Protein Atlas uploads images of these targets ID?\nAlso adding images for targets IDs: 9 and 27 will be great for a better accuracy.\n\nBest Regards,\nJulian Vidal",
    "403137": "Probably, you read only first class in list of classes, because all images are in, e.g. image 01314a58-bba3-11e8-b2b9-ac1f6b6435d0 has 10th class. Also, you can see correct distribution in others kernels (e.g. https://www.kaggle.com/allunia/protein-atlas-exporation-and-baseline).",
    "404314": "I checked and I have no empty class. \n\nNote that some images have 2 or 3 classes (maybe more)\n\nYes, I agree we should have more images for certain classes, especially 27, which is really really lacking.",
    "405060": "I mean that there are no images that ONLY have 1 target ID 10, 15 and 17. I mean, e.g 5dbfaf28-bba6-11e8-b2ba-ac1f6b6435d0 has target ID 10 and 9. but there is no any image with ONLY ID target 10",
    "405061": "Can you please give the name of an image that ONLY has ID target 10, 15 or 17? Thanks a lot!",
    "405160": "I didn't calculate that, but I'm sure there are images with those IDs, and they're more than images with ID 27.\n\nIf we are supposing it right about your table (you got the first ID instead of all IDs), there isn't an image with \"only\" one of those classes.",
    "407952": "It is now proven that you could not use this as a rule to \"post-process\" results. As [LB probing][1] points out, there are classes in LB only have 17. But 10, 15 cannot be certain because of the round down error. Hope it helps. :)\n\n\n  [1]: https://www.kaggle.com/c/human-protein-atlas-image-classification/discussion/68678"
  },
  "source": "meta"
}