{
  "id": 92342,
  "title": "Dataset",
  "url": "/competitions/cassava-disease/discussion/92342",
  "author_name": "",
  "post_date": "2019-05-15T14:28:37.680452800Z",
  "votes": 2,
  "comment_count": 11,
  "views": 0,
  "content": "<p>Why it is mentioned as the dataset consists of 9,436 annotated images while only 5,656 labeled images are available?</p>",
  "messages": [
    {
      "id": "531777",
      "postDate": "05/15/2019 14:28:37",
      "content": "<p>Why it is mentioned as the dataset consists of 9,436 annotated images while only 5,656 labeled images are available?</p>",
      "rawMarkdown": "Why it is mentioned as the dataset consists of 9,436 annotated images while only 5,656 labeled images are available?",
      "votes": null
    },
    {
      "id": "534069",
      "postDate": "05/20/2019 15:26:47",
      "content": "<p>the train and test sets contain 9,436 labeled images in total</p>",
      "rawMarkdown": "the train and test sets contain 9,436 labeled images in total",
      "votes": null
    },
    {
      "id": "534087",
      "postDate": "05/20/2019 16:40:19",
      "content": "<p><a href=\"/tgebru\">@tgebru</a> The test sets are not labeled, right? </p>",
      "rawMarkdown": "tgebru The test sets are not labeled, right?",
      "votes": null
    },
    {
      "id": "534090",
      "postDate": "05/20/2019 16:47:50",
      "content": "<p>They have been labeled by the annotators but you don't see the labels. Where as the additional unlabeled images we make available for training have not been annotated at all.</p>",
      "rawMarkdown": "They have been labeled by the annotators but you don't see the labels. Where as the additional unlabeled images we make available for training have not been annotated at all.",
      "votes": null
    },
    {
      "id": "534096",
      "postDate": "05/20/2019 17:02:38",
      "content": "<p>I am still confused.  For clarification, is this accurate?\n<code>Train Folder = Labeled Images</code>\n<code>Test Folder = Unlabeled Images</code></p>",
      "rawMarkdown": "I am still confused.  For clarification, is this accurate?\n`Train Folder = Labeled Images`\n`Test Folder = Unlabeled Images`",
      "votes": null
    },
    {
      "id": "534097",
      "postDate": "05/20/2019 17:13:51",
      "content": "<p>Yes. You don't have the labels fo the test set. We do.</p>",
      "rawMarkdown": "Yes. You don't have the labels fo the test set. We do.",
      "votes": null
    },
    {
      "id": "534100",
      "postDate": "05/20/2019 17:18:21",
      "content": "<p>I see.  In general, if I want to use these images for training, what's a good method of labeling them?  I have heard of a few tools that can group images to jump start the manual labeling process, but I am curious what a good industry standard approach would be.  </p>",
      "rawMarkdown": "I see.  In general, if I want to use these images for training, what's a good method of labeling them?  I have heard of a few tools that can group images to jump start the manual labeling process, but I am curious what a good industry standard approach would be.",
      "votes": null
    },
    {
      "id": "534102",
      "postDate": "05/20/2019 17:22:23",
      "content": "<p>I don't know of any industry standard but label propagation and semi supervised learning are areas of research that you can look into. </p>",
      "rawMarkdown": "I don't know of any industry standard but label propagation and semi supervised learning are areas of research that you can look into.",
      "votes": null
    },
    {
      "id": "534103",
      "postDate": "05/20/2019 17:22:46",
      "content": "<p>But you should not use the test set during training of course. </p>",
      "rawMarkdown": "But you should not use the test set during training of course.",
      "votes": null
    },
    {
      "id": "534135",
      "postDate": "05/20/2019 18:28:58",
      "content": "<p><a href=\"/tgebru\">@tgebru</a> <code>But you should not use the test set during training of course.</code> - Right, that makes sense.  The extra unlabelled images are in a folder called <code>extraimages</code>.  </p>",
      "rawMarkdown": "tgebru `But you should not use the test set during training of course.` - Right, that makes sense.  The extra unlabelled images are in a folder called `extraimages`.",
      "votes": null
    },
    {
      "id": "536927",
      "postDate": "05/25/2019 17:26:50",
      "content": "",
      "rawMarkdown": "",
      "votes": null
    },
    {
      "id": "538090",
      "postDate": "05/28/2019 05:54:36",
      "content": "<p>Thank you</p>",
      "rawMarkdown": "Thank you",
      "votes": null
    }
  ],
  "comments": [
    {
      "id": 534069,
      "author_name": "tgebru",
      "author_url": "",
      "post_date": "05/20/2019 15:26:47",
      "content": "<p>the train and test sets contain 9,436 labeled images in total</p>",
      "votes": null,
      "replies": [
        {
          "id": 538090,
          "author_name": "abdela",
          "author_url": "",
          "post_date": "05/28/2019 05:54:36",
          "content": "<p>Thank you</p>",
          "votes": null,
          "replies": []
        }
      ]
    },
    {
      "id": 534087,
      "author_name": "jedhassell",
      "author_url": "",
      "post_date": "05/20/2019 16:40:19",
      "content": "<p><a href=\"/tgebru\">@tgebru</a> The test sets are not labeled, right? </p>",
      "votes": null,
      "replies": [
        {
          "id": 536927,
          "author_name": "abdela",
          "author_url": "",
          "post_date": "05/25/2019 17:26:50",
          "content": "",
          "votes": null,
          "replies": []
        }
      ]
    },
    {
      "id": 534090,
      "author_name": "tgebru",
      "author_url": "",
      "post_date": "05/20/2019 16:47:50",
      "content": "<p>They have been labeled by the annotators but you don't see the labels. Where as the additional unlabeled images we make available for training have not been annotated at all.</p>",
      "votes": null,
      "replies": [
        {
          "id": 534096,
          "author_name": "jedhassell",
          "author_url": "",
          "post_date": "05/20/2019 17:02:38",
          "content": "<p>I am still confused.  For clarification, is this accurate?\n<code>Train Folder = Labeled Images</code>\n<code>Test Folder = Unlabeled Images</code></p>",
          "votes": null,
          "replies": []
        }
      ]
    },
    {
      "id": 534097,
      "author_name": "tgebru",
      "author_url": "",
      "post_date": "05/20/2019 17:13:51",
      "content": "<p>Yes. You don't have the labels fo the test set. We do.</p>",
      "votes": null,
      "replies": []
    },
    {
      "id": 534100,
      "author_name": "jedhassell",
      "author_url": "",
      "post_date": "05/20/2019 17:18:21",
      "content": "<p>I see.  In general, if I want to use these images for training, what's a good method of labeling them?  I have heard of a few tools that can group images to jump start the manual labeling process, but I am curious what a good industry standard approach would be.  </p>",
      "votes": null,
      "replies": []
    },
    {
      "id": 534102,
      "author_name": "tgebru",
      "author_url": "",
      "post_date": "05/20/2019 17:22:23",
      "content": "<p>I don't know of any industry standard but label propagation and semi supervised learning are areas of research that you can look into. </p>",
      "votes": null,
      "replies": []
    },
    {
      "id": 534103,
      "author_name": "tgebru",
      "author_url": "",
      "post_date": "05/20/2019 17:22:46",
      "content": "<p>But you should not use the test set during training of course. </p>",
      "votes": null,
      "replies": []
    },
    {
      "id": 534135,
      "author_name": "jedhassell",
      "author_url": "",
      "post_date": "05/20/2019 18:28:58",
      "content": "<p><a href=\"/tgebru\">@tgebru</a> <code>But you should not use the test set during training of course.</code> - Right, that makes sense.  The extra unlabelled images are in a folder called <code>extraimages</code>.  </p>",
      "votes": null,
      "replies": []
    }
  ],
  "raw_markdown_by_id": {
    "531777": "Why it is mentioned as the dataset consists of 9,436 annotated images while only 5,656 labeled images are available?",
    "534069": "the train and test sets contain 9,436 labeled images in total",
    "534087": "tgebru The test sets are not labeled, right?",
    "534090": "They have been labeled by the annotators but you don't see the labels. Where as the additional unlabeled images we make available for training have not been annotated at all.",
    "534096": "I am still confused.  For clarification, is this accurate?\n`Train Folder = Labeled Images`\n`Test Folder = Unlabeled Images`",
    "534097": "Yes. You don't have the labels fo the test set. We do.",
    "534100": "I see.  In general, if I want to use these images for training, what's a good method of labeling them?  I have heard of a few tools that can group images to jump start the manual labeling process, but I am curious what a good industry standard approach would be.",
    "534102": "I don't know of any industry standard but label propagation and semi supervised learning are areas of research that you can look into.",
    "534103": "But you should not use the test set during training of course.",
    "534135": "tgebru `But you should not use the test set during training of course.` - Right, that makes sense.  The extra unlabelled images are in a folder called `extraimages`.",
    "536927": "",
    "538090": "Thank you"
  },
  "source": "meta"
}