{
  "id": 253475,
  "title": "Exact number of classes we're to predict on",
  "url": "/competitions/siim-covid19-detection/discussion/253475",
  "author_name": "",
  "post_date": "2021-07-16T20:16:07.327395Z",
  "votes": null,
  "comment_count": 1,
  "views": 0,
  "content": "<p>Hi, reading over the overview and the data pages, for study label labels  we're predicting for 4 classes while for image level labels we're predicting on 2 labels.. My question is are we predicting all 6 labels with one model or we're predicting only the study labels alone?</p>\n<p>Also can someone explain the why the test set differs from the number of observations in the sample_submission.csv?</p>",
  "messages": [
    {
      "id": "1390596",
      "postDate": "07/16/2021 20:16:07",
      "content": "<p>Hi, reading over the overview and the data pages, for study label labels  we're predicting for 4 classes while for image level labels we're predicting on 2 labels.. My question is are we predicting all 6 labels with one model or we're predicting only the study labels alone?</p>\n<p>Also can someone explain the why the test set differs from the number of observations in the sample_submission.csv?</p>",
      "rawMarkdown": "Hi, reading over the overview and the data pages, for study label labels  we're predicting for 4 classes while for image level labels we're predicting on 2 labels.. My question is are we predicting all 6 labels with one model or we're predicting only the study labels alone?\n\nAlso can someone explain the why the test set differs from the number of observations in the sample_submission.csv?",
      "votes": null
    },
    {
      "id": "1398861",
      "postDate": "07/24/2021 15:14:17",
      "content": "<p>if you look at the dataset, the study label is linked to image level. For each image, you need to perform 2 class label (none or opacity). Some each study, you need to predict the 4 class. When you look at the sample submission file, there are 2 type of label, one end with _image while the other end with _sutdy. <br>\nFor those end with _image is the one you need to do the 2 class label, followed by conf and bbox. for the one end with _study, you need to perform 2 class label with conf score and end with 0 0 1 1. </p>\n<p>Some study have linked to mutiple images. </p>",
      "rawMarkdown": "if you look at the dataset, the study label is linked to image level. For each image, you need to perform 2 class label (none or opacity). Some each study, you need to predict the 4 class. When you look at the sample submission file, there are 2 type of label, one end with _image while the other end with _sutdy. \nFor those end with _image is the one you need to do the 2 class label, followed by conf and bbox. for the one end with _study, you need to perform 2 class label with conf score and end with 0 0 1 1. \n\nSome study have linked to mutiple images.",
      "votes": null
    }
  ],
  "comments": [
    {
      "id": 1398861,
      "author_name": "hawkeat",
      "author_url": "",
      "post_date": "07/24/2021 15:14:17",
      "content": "<p>if you look at the dataset, the study label is linked to image level. For each image, you need to perform 2 class label (none or opacity). Some each study, you need to predict the 4 class. When you look at the sample submission file, there are 2 type of label, one end with _image while the other end with _sutdy. <br>\nFor those end with _image is the one you need to do the 2 class label, followed by conf and bbox. for the one end with _study, you need to perform 2 class label with conf score and end with 0 0 1 1. </p>\n<p>Some study have linked to mutiple images. </p>",
      "votes": null,
      "replies": []
    }
  ],
  "raw_markdown_by_id": {
    "1390596": "Hi, reading over the overview and the data pages, for study label labels  we're predicting for 4 classes while for image level labels we're predicting on 2 labels.. My question is are we predicting all 6 labels with one model or we're predicting only the study labels alone?\n\nAlso can someone explain the why the test set differs from the number of observations in the sample_submission.csv?",
    "1398861": "if you look at the dataset, the study label is linked to image level. For each image, you need to perform 2 class label (none or opacity). Some each study, you need to predict the 4 class. When you look at the sample submission file, there are 2 type of label, one end with _image while the other end with _sutdy. \nFor those end with _image is the one you need to do the 2 class label, followed by conf and bbox. for the one end with _study, you need to perform 2 class label with conf score and end with 0 0 1 1. \n\nSome study have linked to mutiple images."
  },
  "source": "meta"
}