{
  "id": 64389,
  "title": "Clever evaluation metric",
  "url": "/competitions/rsna-pneumonia-detection-challenge/discussion/64389",
  "author_name": "",
  "post_date": "2018-08-29T04:03:57.522189700Z",
  "votes": 1,
  "comment_count": 1,
  "views": 0,
  "content": "<p>I just want to comment about the evaluation metric used. Globally, it looks like a clever evaluation metric to average the binary classification performance and the detection bounding box precision needed for the global task. </p>\n\n<p>This metric could scientifically be used more frequently to evaluate detection models on unbalanced detection datasets with a lot of negative samples. This situation is encountered very frequently in many radiology problems.</p>\n\n<p>Even if hidden behind the detection task, I think the binary classification task will be the tougher challenge overall.</p>\n\n<p><strong>Any thoughts anyone about the evaluation metric ?</strong> </p>",
  "messages": [
    {
      "id": "377378",
      "postDate": "08/29/2018 04:03:57",
      "content": "<p>I just want to comment about the evaluation metric used. Globally, it looks like a clever evaluation metric to average the binary classification performance and the detection bounding box precision needed for the global task. </p>\n\n<p>This metric could scientifically be used more frequently to evaluate detection models on unbalanced detection datasets with a lot of negative samples. This situation is encountered very frequently in many radiology problems.</p>\n\n<p>Even if hidden behind the detection task, I think the binary classification task will be the tougher challenge overall.</p>\n\n<p><strong>Any thoughts anyone about the evaluation metric ?</strong> </p>",
      "rawMarkdown": "I just want to comment about the evaluation metric used. Globally, it looks like a clever evaluation metric to average the binary classification performance and the detection bounding box precision needed for the global task. \n\nThis metric could scientifically be used more frequently to evaluate detection models on unbalanced detection datasets with a lot of negative samples. This situation is encountered very frequently in many radiology problems.\n\nEven if hidden behind the detection task, I think the binary classification task will be the tougher challenge overall.\n\n**Any thoughts anyone about the evaluation metric ?**",
      "votes": null
    },
    {
      "id": "380178",
      "postDate": "09/01/2018 22:07:58",
      "content": "<p>Actually there are some questions arises then you try to make clear the evaluation procedure. Please join the discussion - <a href=\"https://www.kaggle.com/c/rsna-pneumonia-detection-challenge/discussion/64542\">https://www.kaggle.com/c/rsna-pneumonia-detection-challenge/discussion/64542</a></p>\n\n<p>May you help us to this?</p>",
      "rawMarkdown": "Actually there are some questions arises then you try to make clear the evaluation procedure. Please join the discussion - https://www.kaggle.com/c/rsna-pneumonia-detection-challenge/discussion/64542\n\nMay you help us to this?",
      "votes": null
    }
  ],
  "comments": [
    {
      "id": 380178,
      "author_name": "ivanpozd",
      "author_url": "",
      "post_date": "09/01/2018 22:07:58",
      "content": "<p>Actually there are some questions arises then you try to make clear the evaluation procedure. Please join the discussion - <a href=\"https://www.kaggle.com/c/rsna-pneumonia-detection-challenge/discussion/64542\">https://www.kaggle.com/c/rsna-pneumonia-detection-challenge/discussion/64542</a></p>\n\n<p>May you help us to this?</p>",
      "votes": null,
      "replies": []
    }
  ],
  "raw_markdown_by_id": {
    "377378": "I just want to comment about the evaluation metric used. Globally, it looks like a clever evaluation metric to average the binary classification performance and the detection bounding box precision needed for the global task. \n\nThis metric could scientifically be used more frequently to evaluate detection models on unbalanced detection datasets with a lot of negative samples. This situation is encountered very frequently in many radiology problems.\n\nEven if hidden behind the detection task, I think the binary classification task will be the tougher challenge overall.\n\n**Any thoughts anyone about the evaluation metric ?**",
    "380178": "Actually there are some questions arises then you try to make clear the evaluation procedure. Please join the discussion - https://www.kaggle.com/c/rsna-pneumonia-detection-challenge/discussion/64542\n\nMay you help us to this?"
  },
  "source": "meta"
}