{
  "id": 318114,
  "title": "mAP Score Calculation + Meaning",
  "url": "/competitions/siim-covid19-detection/discussion/318114",
  "author_name": "Bill Xu",
  "post_date": "2022-04-10T17:24:40.791000",
  "votes": 0,
  "comment_count": 0,
  "views": 0,
  "content": "<p>Could someone please explain to me how the mAP score for this competition is calculated? From what I have found online, it is the area under the recall/precision graph averaged over all the classes. But, is this average weighted in some way? And also how does the accuracy of the classification model weigh into the mAP score? </p>\n<p>If I had an mAP score of 0.6, what would it mean? It wouldn't mean that I predicted 60 percent of the images correctly, right? My other thought was that it meant the models together predicted 60 percent of the bounding boxes and classifications correctly, but that doesn't seem right either. </p>\n<p>Thanks in advance.</p>",
  "messages": [
    {
      "id": 1751375,
      "postDate": "2022-04-10T17:24:40.790Z",
      "content": "<p>Could someone please explain to me how the mAP score for this competition is calculated? From what I have found online, it is the area under the recall/precision graph averaged over all the classes. But, is this average weighted in some way? And also how does the accuracy of the classification model weigh into the mAP score? </p>\n<p>If I had an mAP score of 0.6, what would it mean? It wouldn't mean that I predicted 60 percent of the images correctly, right? My other thought was that it meant the models together predicted 60 percent of the bounding boxes and classifications correctly, but that doesn't seem right either. </p>\n<p>Thanks in advance.</p>",
      "rawMarkdown": "Could someone please explain to me how the mAP score for this competition is calculated? From what I have found online, it is the area under the recall/precision graph averaged over all the classes. But, is this average weighted in some way? And also how does the accuracy of the classification model weigh into the mAP score? \n\nIf I had an mAP score of 0.6, what would it mean? It wouldn't mean that I predicted 60 percent of the images correctly, right? My other thought was that it meant the models together predicted 60 percent of the bounding boxes and classifications correctly, but that doesn't seem right either. \n\nThanks in advance."
    }
  ],
  "comments": [],
  "raw_markdown_by_id": {
    "1751375": "Could someone please explain to me how the mAP score for this competition is calculated? From what I have found online, it is the area under the recall/precision graph averaged over all the classes. But, is this average weighted in some way? And also how does the accuracy of the classification model weigh into the mAP score? \n\nIf I had an mAP score of 0.6, what would it mean? It wouldn't mean that I predicted 60 percent of the images correctly, right? My other thought was that it meant the models together predicted 60 percent of the bounding boxes and classifications correctly, but that doesn't seem right either. \n\nThanks in advance."
  }
}