{
  "id": 351177,
  "title": "How to ensemble?",
  "url": "/competitions/hubmap-organ-segmentation/discussion/351177",
  "author_name": "Jiseong Ok",
  "post_date": "2022-09-08T20:34:55.029000",
  "votes": 3,
  "comment_count": 3,
  "views": 0,
  "content": "<p>I'm just using mean value with multiple model weights.<br>\nAre there effective methods of ensemble in segmentation?🙏</p>",
  "messages": [
    {
      "id": 1931596,
      "postDate": "2022-09-08T20:34:55.030Z",
      "content": "<p>I'm just using mean value with multiple model weights.<br>\nAre there effective methods of ensemble in segmentation?🙏</p>",
      "rawMarkdown": "I'm just using mean value with multiple model weights.\nAre there effective methods of ensemble in segmentation?🙏",
      "votes": 2
    },
    {
      "id": 1934010,
      "postDate": "2022-09-11T04:12:13.727Z",
      "content": "<p>I used plain mean of predictions from different folds, then get mask above a threshold. <a href=\"https://www.kaggle.com/jiseongok\" target=\"_blank\">@jiseongok</a> How do you determine model weights?</p>",
      "rawMarkdown": "I used plain mean of predictions from different folds, then get mask above a threshold. @jiseongok How do you determine model weights?",
      "replies": [
        {
          "id": 1935265,
          "postDate": "2022-09-12T01:45:37.420Z",
          "rawMarkdown": "",
          "isDeleted": true
        },
        {
          "id": 1935266,
          "postDate": "2022-09-12T01:47:13.063Z",
          "content": "<p>I used various tile set. For example, 256-12, 256-6 and 256-4 tile set. And ensemble.</p>",
          "rawMarkdown": "I used various tile set. For example, 256-12, 256-6 and 256-4 tile set. And ensemble.",
          "votes": 2
        }
      ]
    }
  ],
  "comments": [
    {
      "id": 1934010,
      "author_name": "Xiang Zhang",
      "author_url": "",
      "post_date": "2022-09-11T04:12:13.727000",
      "content": "<p>I used plain mean of predictions from different folds, then get mask above a threshold. <a href=\"https://www.kaggle.com/jiseongok\" target=\"_blank\">@jiseongok</a> How do you determine model weights?</p>",
      "votes": 0,
      "replies": [
        {
          "id": 1935265,
          "author_name": "",
          "author_url": "",
          "post_date": "2022-09-12T01:45:37.420000",
          "content": "",
          "votes": 0,
          "replies": []
        },
        {
          "id": 1935266,
          "author_name": "Jiseong Ok",
          "author_url": "",
          "post_date": "2022-09-12T01:47:13.063000",
          "content": "<p>I used various tile set. For example, 256-12, 256-6 and 256-4 tile set. And ensemble.</p>",
          "votes": 2,
          "replies": []
        }
      ]
    }
  ],
  "raw_markdown_by_id": {
    "1931596": "I'm just using mean value with multiple model weights.\nAre there effective methods of ensemble in segmentation?🙏",
    "1934010": "I used plain mean of predictions from different folds, then get mask above a threshold. @jiseongok How do you determine model weights?"
  }
}