{
  "id": 419536,
  "title": "How to apply semi-supervised learning to instance segmentation tasks?",
  "url": "/competitions/hubmap-hacking-the-human-vasculature/discussion/419536",
  "author_name": "",
  "post_date": "2023-06-26T12:04:55.269026700Z",
  "votes": 1,
  "comment_count": 4,
  "views": 0,
  "content": "<p>any ideas/papers/example code?</p>",
  "messages": [
    {
      "id": "2318489",
      "postDate": "06/26/2023 12:04:55",
      "content": "<p>any ideas/papers/example code?</p>",
      "rawMarkdown": "any ideas/papers/example code?",
      "votes": null
    },
    {
      "id": "2318590",
      "postDate": "06/26/2023 13:08:53",
      "content": "<p>Why would you do semi-supervised when we have labels?</p>",
      "rawMarkdown": "Why would you do semi-supervised when we have labels?",
      "votes": null
    },
    {
      "id": "2318602",
      "postDate": "06/26/2023 13:21:29",
      "content": "<p>Well, as the Data Description says: <code>We also include, as Dataset 3, tiles extracted from an additional nine WSIs. These tiles have not been annotated. You may wish to apply semi- or self-supervised learning techniques on this data to support your predictions.</code></p>",
      "rawMarkdown": "Well, as the Data Description says: `We also include, as Dataset 3, tiles extracted from an additional nine WSIs. These tiles have not been annotated. You may wish to apply semi- or self-supervised learning techniques on this data to support your predictions.`",
      "votes": null
    },
    {
      "id": "2318678",
      "postDate": "06/26/2023 14:04:36",
      "content": "<p>Right I forgot about that part of the dataset. I think you can predict instance masks for those unannotated WSIs by using an ensemble and then concatenate them to your training dataset.</p>",
      "rawMarkdown": "Right I forgot about that part of the dataset. I think you can predict instance masks for those unannotated WSIs by using an ensemble and then concatenate them to your training dataset.",
      "votes": null
    },
    {
      "id": "2319174",
      "postDate": "06/27/2023 00:31:57",
      "content": "<p>Okay, I'll definitely give serious thought. Thanks a lot!</p>",
      "rawMarkdown": "Okay, I'll definitely give serious thought. Thanks a lot!",
      "votes": null
    }
  ],
  "comments": [
    {
      "id": 2318590,
      "author_name": "gunesevitan",
      "author_url": "",
      "post_date": "06/26/2023 13:08:53",
      "content": "<p>Why would you do semi-supervised when we have labels?</p>",
      "votes": null,
      "replies": [
        {
          "id": 2318602,
          "author_name": "wqx20000115",
          "author_url": "",
          "post_date": "06/26/2023 13:21:29",
          "content": "<p>Well, as the Data Description says: <code>We also include, as Dataset 3, tiles extracted from an additional nine WSIs. These tiles have not been annotated. You may wish to apply semi- or self-supervised learning techniques on this data to support your predictions.</code></p>",
          "votes": null,
          "replies": [
            {
              "id": 2318678,
              "author_name": "gunesevitan",
              "author_url": "",
              "post_date": "06/26/2023 14:04:36",
              "content": "<p>Right I forgot about that part of the dataset. I think you can predict instance masks for those unannotated WSIs by using an ensemble and then concatenate them to your training dataset.</p>",
              "votes": null,
              "replies": [
                {
                  "id": 2319174,
                  "author_name": "wqx20000115",
                  "author_url": "",
                  "post_date": "06/27/2023 00:31:57",
                  "content": "<p>Okay, I'll definitely give serious thought. Thanks a lot!</p>",
                  "votes": null,
                  "replies": []
                }
              ]
            }
          ]
        }
      ]
    }
  ],
  "raw_markdown_by_id": {
    "2318489": "any ideas/papers/example code?",
    "2318590": "Why would you do semi-supervised when we have labels?",
    "2318602": "Well, as the Data Description says: `We also include, as Dataset 3, tiles extracted from an additional nine WSIs. These tiles have not been annotated. You may wish to apply semi- or self-supervised learning techniques on this data to support your predictions.`",
    "2318678": "Right I forgot about that part of the dataset. I think you can predict instance masks for those unannotated WSIs by using an ensemble and then concatenate them to your training dataset.",
    "2319174": "Okay, I'll definitely give serious thought. Thanks a lot!"
  },
  "source": "meta"
}