{
  "id": 251807,
  "title": "Multiple Images for Study in Test Folder",
  "url": "/competitions/siim-covid19-detection/discussion/251807",
  "author_name": "",
  "post_date": "2021-07-08T22:50:57.617933100Z",
  "votes": 1,
  "comment_count": 2,
  "views": 0,
  "content": "<p>I have a question: <br>\n   If we have multiple images for a study (when a study contains two or more different subfolders, where each subfolder has one or more images (e.x., '342624fb2b11' study in test folder has 4 substudies, where 3 of them have 2 images)), do we predict study labels (i.e. \"negative\", \"typical\", \"indeterminate\", \"atypical\") for each image in subfolders, take mean of these predictions and submit the result (mean) as the final prediction for the whole study?</p>",
  "messages": [
    {
      "id": "1381365",
      "postDate": "07/08/2021 22:50:57",
      "content": "<p>I have a question: <br>\n   If we have multiple images for a study (when a study contains two or more different subfolders, where each subfolder has one or more images (e.x., '342624fb2b11' study in test folder has 4 substudies, where 3 of them have 2 images)), do we predict study labels (i.e. \"negative\", \"typical\", \"indeterminate\", \"atypical\") for each image in subfolders, take mean of these predictions and submit the result (mean) as the final prediction for the whole study?</p>",
      "rawMarkdown": "I have a question: \n   If we have multiple images for a study (when a study contains two or more different subfolders, where each subfolder has one or more images (e.x., '342624fb2b11' study in test folder has 4 substudies, where 3 of them have 2 images)), do we predict study labels (i.e. \"negative\", \"typical\", \"indeterminate\", \"atypical\") for each image in subfolders, take mean of these predictions and submit the result (mean) as the final prediction for the whole study?",
      "votes": null
    },
    {
      "id": "1382224",
      "postDate": "07/09/2021 15:49:33",
      "content": "<p>I would assume that's what you do. That's what I did. Either way I think you just need to predict a study label. How you do that is up to you, you could also just pick one image randomly and make your prediction based off of it, but I think deciding your labels based off of the mode would be best.</p>",
      "rawMarkdown": "I would assume that's what you do. That's what I did. Either way I think you just need to predict a study label. How you do that is up to you, you could also just pick one image randomly and make your prediction based off of it, but I think deciding your labels based off of the mode would be best.",
      "votes": null
    },
    {
      "id": "1382277",
      "postDate": "07/09/2021 16:44:20",
      "content": "<p>Yeah, I think so too. Thank you for your reply</p>",
      "rawMarkdown": "Yeah, I think so too. Thank you for your reply",
      "votes": null
    }
  ],
  "comments": [
    {
      "id": 1382224,
      "author_name": "rex1545",
      "author_url": "",
      "post_date": "07/09/2021 15:49:33",
      "content": "<p>I would assume that's what you do. That's what I did. Either way I think you just need to predict a study label. How you do that is up to you, you could also just pick one image randomly and make your prediction based off of it, but I think deciding your labels based off of the mode would be best.</p>",
      "votes": null,
      "replies": [
        {
          "id": 1382277,
          "author_name": "antonzv",
          "author_url": "",
          "post_date": "07/09/2021 16:44:20",
          "content": "<p>Yeah, I think so too. Thank you for your reply</p>",
          "votes": null,
          "replies": []
        }
      ]
    }
  ],
  "raw_markdown_by_id": {
    "1381365": "I have a question: \n   If we have multiple images for a study (when a study contains two or more different subfolders, where each subfolder has one or more images (e.x., '342624fb2b11' study in test folder has 4 substudies, where 3 of them have 2 images)), do we predict study labels (i.e. \"negative\", \"typical\", \"indeterminate\", \"atypical\") for each image in subfolders, take mean of these predictions and submit the result (mean) as the final prediction for the whole study?",
    "1382224": "I would assume that's what you do. That's what I did. Either way I think you just need to predict a study label. How you do that is up to you, you could also just pick one image randomly and make your prediction based off of it, but I think deciding your labels based off of the mode would be best.",
    "1382277": "Yeah, I think so too. Thank you for your reply"
  },
  "source": "meta"
}