{
  "id": 262804,
  "title": "Has anyone been calculating mAP using torchmetrics?",
  "url": "/competitions/siim-covid19-detection/discussion/262804",
  "author_name": "",
  "post_date": "2021-08-07T15:04:02.658568100Z",
  "votes": 2,
  "comment_count": 2,
  "views": 0,
  "content": "<p>On study prediction, I saw average_precision_score from sklearn has been adopted across the competition (people are mentioning it in Discussion) most of the time.</p>\n<p>For myself, I have been using torchmetrics from the start because it is easy to use and fast <br>\nFrom my experience, my CV that calculated using torchmetrics almost identical to LB most of the time. I do noticed that this might be irrelevant since it depends on how good model performing on test dataset.</p>\n<p>I just wondered if there was any significant difference between two of those.</p>",
  "messages": [
    {
      "id": "1457786",
      "postDate": "08/07/2021 15:04:02",
      "content": "<p>On study prediction, I saw average_precision_score from sklearn has been adopted across the competition (people are mentioning it in Discussion) most of the time.</p>\n<p>For myself, I have been using torchmetrics from the start because it is easy to use and fast <br>\nFrom my experience, my CV that calculated using torchmetrics almost identical to LB most of the time. I do noticed that this might be irrelevant since it depends on how good model performing on test dataset.</p>\n<p>I just wondered if there was any significant difference between two of those.</p>",
      "rawMarkdown": "On study prediction, I saw average_precision_score from sklearn has been adopted across the competition (people are mentioning it in Discussion) most of the time.\n\nFor myself, I have been using torchmetrics from the start because it is easy to use and fast \nFrom my experience, my CV that calculated using torchmetrics almost identical to LB most of the time. I do noticed that this might be irrelevant since it depends on how good model performing on test dataset.\n\nI just wondered if there was any significant difference between two of those.",
      "votes": null
    },
    {
      "id": "1458075",
      "postDate": "08/07/2021 17:38:51",
      "content": "<p>There is no difference between the torchmterics and sklearn implementation of mAP. You just need to ensure that the predictions and the labels are being feed in with proper format.</p>",
      "rawMarkdown": "There is no difference between the torchmterics and sklearn implementation of mAP. You just need to ensure that the predictions and the labels are being feed in with proper format.",
      "votes": null
    },
    {
      "id": "1458942",
      "postDate": "08/08/2021 06:26:41",
      "content": "<p>can you share a link to TorchMetrics mAP, please?</p>",
      "rawMarkdown": "can you share a link to TorchMetrics mAP, please?",
      "votes": null
    },
    {
      "id": "1458957",
      "postDate": "08/08/2021 06:35:45",
      "content": "<p><a href=\"https://torchmetrics.readthedocs.io/en/stable/references/modules.html#averageprecision\" target=\"_blank\">https://torchmetrics.readthedocs.io/en/stable/references/modules.html#averageprecision</a></p>\n<p>It used to be part of pytorch lightning now it is standalone package<br>\nAccuracy and Auc also available</p>",
      "rawMarkdown": "https://torchmetrics.readthedocs.io/en/stable/references/modules.html#averageprecision\n\nIt used to be part of pytorch lightning now it is standalone package\nAccuracy and Auc also available",
      "votes": null
    }
  ],
  "comments": [
    {
      "id": 1458075,
      "author_name": "benihime91",
      "author_url": "",
      "post_date": "08/07/2021 17:38:51",
      "content": "<p>There is no difference between the torchmterics and sklearn implementation of mAP. You just need to ensure that the predictions and the labels are being feed in with proper format.</p>",
      "votes": null,
      "replies": []
    },
    {
      "id": 1458942,
      "author_name": "joven1997",
      "author_url": "",
      "post_date": "08/08/2021 06:26:41",
      "content": "<p>can you share a link to TorchMetrics mAP, please?</p>",
      "votes": null,
      "replies": [
        {
          "id": 1458957,
          "author_name": "sunpnwt12",
          "author_url": "",
          "post_date": "08/08/2021 06:35:45",
          "content": "<p><a href=\"https://torchmetrics.readthedocs.io/en/stable/references/modules.html#averageprecision\" target=\"_blank\">https://torchmetrics.readthedocs.io/en/stable/references/modules.html#averageprecision</a></p>\n<p>It used to be part of pytorch lightning now it is standalone package<br>\nAccuracy and Auc also available</p>",
          "votes": null,
          "replies": []
        }
      ]
    }
  ],
  "raw_markdown_by_id": {
    "1457786": "On study prediction, I saw average_precision_score from sklearn has been adopted across the competition (people are mentioning it in Discussion) most of the time.\n\nFor myself, I have been using torchmetrics from the start because it is easy to use and fast \nFrom my experience, my CV that calculated using torchmetrics almost identical to LB most of the time. I do noticed that this might be irrelevant since it depends on how good model performing on test dataset.\n\nI just wondered if there was any significant difference between two of those.",
    "1458075": "There is no difference between the torchmterics and sklearn implementation of mAP. You just need to ensure that the predictions and the labels are being feed in with proper format.",
    "1458942": "can you share a link to TorchMetrics mAP, please?",
    "1458957": "https://torchmetrics.readthedocs.io/en/stable/references/modules.html#averageprecision\n\nIt used to be part of pytorch lightning now it is standalone package\nAccuracy and Auc also available"
  },
  "source": "meta"
}