{
  "id": 223060,
  "title": "How to use annotation efficiently?",
  "url": "/competitions/ranzcr-clip-catheter-line-classification/discussion/223060",
  "author_name": "Aman Deep Gupta",
  "post_date": "2021-03-02T10:26:57.089000",
  "votes": 1,
  "comment_count": 0,
  "views": 0,
  "content": "<p>Here total images are 30k but annotation are provided only for 9k images. We can use them in teacher student model but is there is any way to predict annotations on rest 21k training images?  And then use them for training our models.<br>\nHow we can get segmentation for the rest images? </p>\n<p>Which is better approach:</p>\n<ol>\n<li>Made a model to predict segmentation and class simultaneously.</li>\n<li>Made one model to predict segmentation and then use this output as input for another model to predict on 11 classes.</li>\n</ol>\n<p>Thanks for your ideas in advance😊</p>",
  "messages": [
    {
      "id": 1222956,
      "postDate": "2021-03-02T10:26:57.090Z",
      "content": "<p>Here total images are 30k but annotation are provided only for 9k images. We can use them in teacher student model but is there is any way to predict annotations on rest 21k training images?  And then use them for training our models.<br>\nHow we can get segmentation for the rest images? </p>\n<p>Which is better approach:</p>\n<ol>\n<li>Made a model to predict segmentation and class simultaneously.</li>\n<li>Made one model to predict segmentation and then use this output as input for another model to predict on 11 classes.</li>\n</ol>\n<p>Thanks for your ideas in advance😊</p>",
      "rawMarkdown": "Here total images are 30k but annotation are provided only for 9k images. We can use them in teacher student model but is there is any way to predict annotations on rest 21k training images?  And then use them for training our models.\nHow we can get segmentation for the rest images? \n\nWhich is better approach:\n1. Made a model to predict segmentation and class simultaneously.\n2. Made one model to predict segmentation and then use this output as input for another model to predict on 11 classes.\n\nThanks for your ideas in advance😊",
      "votes": 1
    }
  ],
  "comments": [],
  "raw_markdown_by_id": {
    "1222956": "Here total images are 30k but annotation are provided only for 9k images. We can use them in teacher student model but is there is any way to predict annotations on rest 21k training images?  And then use them for training our models.\nHow we can get segmentation for the rest images? \n\nWhich is better approach:\n1. Made a model to predict segmentation and class simultaneously.\n2. Made one model to predict segmentation and then use this output as input for another model to predict on 11 classes.\n\nThanks for your ideas in advance😊"
  }
}