{
  "id": 116744,
  "title": "How to train or model such that it predicts if there is no specific cloud type in an image??",
  "url": "/competitions/understanding_cloud_organization/discussion/116744",
  "author_name": "",
  "post_date": "2019-11-11T07:49:23.310024100Z",
  "votes": null,
  "comment_count": 3,
  "views": 0,
  "content": "<p>I am a newbie please help</p>",
  "messages": [
    {
      "id": "670235",
      "postDate": "11/11/2019 07:49:23",
      "content": "<p>I am a newbie please help</p>",
      "rawMarkdown": "I am a newbie please help",
      "votes": null
    },
    {
      "id": "670299",
      "postDate": "11/11/2019 09:39:29",
      "content": "<ol>\n<li>set network output to (batch_size x 4)</li>\n<li>reduce mask (batch_size x 4 x H x W) to (batch_size x 4), e.g. mask.sum(dim=[2,3])&gt;0 in pytorch\nyou can predict whether any one cloud type is present in image</li>\n</ol>",
      "rawMarkdown": "1. set network output to (batch_size x 4)\n2. reduce mask (batch_size x 4 x H x W) to (batch_size x 4), e.g. mask.sum(dim=[2,3])&gt;0 in pytorch\nyou can predict whether any one cloud type is present in image",
      "votes": null
    },
    {
      "id": "670913",
      "postDate": "11/12/2019 03:25:06",
      "content": "<p>I believe that Heng said that a leak in this competition is that every image has at least one label</p>",
      "rawMarkdown": "I believe that Heng said that a leak in this competition is that every image has at least one label",
      "votes": null
    },
    {
      "id": "670960",
      "postDate": "11/12/2019 04:48:42",
      "content": "<p>not exactly a leak as i realized that it is stated in the competition website.\nbut is a  useful fact that we can exploit</p>\n\n<p>\"Data Description\"\n .... Each image has at least one cloud formation, and can possibly contain up to all all four.</p>",
      "rawMarkdown": "not exactly a leak as i realized that it is stated in the competition website.\nbut is a  useful fact that we can exploit\n\n\"Data Description\"\n .... Each image has at least one cloud formation, and can possibly contain up to all all four.",
      "votes": null
    }
  ],
  "comments": [
    {
      "id": 670299,
      "author_name": "niuddd",
      "author_url": "",
      "post_date": "11/11/2019 09:39:29",
      "content": "<ol>\n<li>set network output to (batch_size x 4)</li>\n<li>reduce mask (batch_size x 4 x H x W) to (batch_size x 4), e.g. mask.sum(dim=[2,3])&gt;0 in pytorch\nyou can predict whether any one cloud type is present in image</li>\n</ol>",
      "votes": null,
      "replies": []
    },
    {
      "id": 670913,
      "author_name": "cweed28",
      "author_url": "",
      "post_date": "11/12/2019 03:25:06",
      "content": "<p>I believe that Heng said that a leak in this competition is that every image has at least one label</p>",
      "votes": null,
      "replies": [
        {
          "id": 670960,
          "author_name": "hengck23",
          "author_url": "",
          "post_date": "11/12/2019 04:48:42",
          "content": "<p>not exactly a leak as i realized that it is stated in the competition website.\nbut is a  useful fact that we can exploit</p>\n\n<p>\"Data Description\"\n .... Each image has at least one cloud formation, and can possibly contain up to all all four.</p>",
          "votes": null,
          "replies": []
        }
      ]
    }
  ],
  "raw_markdown_by_id": {
    "670235": "I am a newbie please help",
    "670299": "1. set network output to (batch_size x 4)\n2. reduce mask (batch_size x 4 x H x W) to (batch_size x 4), e.g. mask.sum(dim=[2,3])&gt;0 in pytorch\nyou can predict whether any one cloud type is present in image",
    "670913": "I believe that Heng said that a leak in this competition is that every image has at least one label",
    "670960": "not exactly a leak as i realized that it is stated in the competition website.\nbut is a  useful fact that we can exploit\n\n\"Data Description\"\n .... Each image has at least one cloud formation, and can possibly contain up to all all four."
  },
  "source": "meta"
}