{
  "id": 179904,
  "title": "Metric: Global Average Precision(GAP)",
  "url": "/competitions/landmark-recognition-2020/discussion/179904",
  "author_name": "",
  "post_date": "2020-09-03T08:28:50.052523500Z",
  "votes": 2,
  "comment_count": 3,
  "views": 0,
  "content": "<p>Hello, For those who were struggling to implement GAP metric, they can also use micro average precision score from sklearn. <code>sklearn.metric.average_precision_score</code> . This will allow many competitors to track the metrics they wanted to maximize. <br>\nP.S. open to suggestions and correct me if something is wrong.</p>",
  "messages": [
    {
      "id": "996342",
      "postDate": "09/03/2020 08:28:50",
      "content": "<p>Hello, For those who were struggling to implement GAP metric, they can also use micro average precision score from sklearn. <code>sklearn.metric.average_precision_score</code> . This will allow many competitors to track the metrics they wanted to maximize. <br>\nP.S. open to suggestions and correct me if something is wrong.</p>",
      "rawMarkdown": "Hello, For those who were struggling to implement GAP metric, they can also use micro average precision score from sklearn. `sklearn.metric.average_precision_score` . This will allow many competitors to track the metrics they wanted to maximize. \nP.S. open to suggestions and correct me if something is wrong.",
      "votes": null
    },
    {
      "id": "998607",
      "postDate": "09/04/2020 21:40:17",
      "content": "<p>Hosts have open-sourced the code to compute GAP. <br>\n<a href=\"https://github.com/tensorflow/models/blob/master/research/delf/delf/python/google_landmarks_dataset/compute_recognition_metrics.py\" target=\"_blank\">https://github.com/tensorflow/models/blob/master/research/delf/delf/python/google_landmarks_dataset/compute_recognition_metrics.py</a></p>",
      "rawMarkdown": "Hosts have open-sourced the code to compute GAP. \nhttps://github.com/tensorflow/models/blob/master/research/delf/delf/python/google_landmarks_dataset/compute_recognition_metrics.py",
      "votes": null
    },
    {
      "id": "1002633",
      "postDate": "09/08/2020 09:44:33",
      "content": "<p>What should be the range of GAP?</p>\n<p>Should it be between 0 and 1 or it is also possible that it's value is &gt;1?</p>\n<p>Asking because for some of the predictions from baseline model are showing values to be &gt;1</p>",
      "rawMarkdown": "What should be the range of GAP?\n\nShould it be between 0 and 1 or it is also possible that it's value is >1?\n\nAsking because for some of the predictions from baseline model are showing values to be >1",
      "votes": null
    },
    {
      "id": "1002698",
      "postDate": "09/08/2020 11:02:33",
      "content": "<p>Confidence scores could be anything, because they are used only for sorting your predictions. So, it can be larger than 1. It actually could be any data type that could be sorted.</p>",
      "rawMarkdown": "Confidence scores could be anything, because they are used only for sorting your predictions. So, it can be larger than 1. It actually could be any data type that could be sorted.",
      "votes": null
    }
  ],
  "comments": [
    {
      "id": 998607,
      "author_name": "rarun2596",
      "author_url": "",
      "post_date": "09/04/2020 21:40:17",
      "content": "<p>Hosts have open-sourced the code to compute GAP. <br>\n<a href=\"https://github.com/tensorflow/models/blob/master/research/delf/delf/python/google_landmarks_dataset/compute_recognition_metrics.py\" target=\"_blank\">https://github.com/tensorflow/models/blob/master/research/delf/delf/python/google_landmarks_dataset/compute_recognition_metrics.py</a></p>",
      "votes": null,
      "replies": []
    },
    {
      "id": 1002633,
      "author_name": "rohitdeepu17",
      "author_url": "",
      "post_date": "09/08/2020 09:44:33",
      "content": "<p>What should be the range of GAP?</p>\n<p>Should it be between 0 and 1 or it is also possible that it's value is &gt;1?</p>\n<p>Asking because for some of the predictions from baseline model are showing values to be &gt;1</p>",
      "votes": null,
      "replies": [
        {
          "id": 1002698,
          "author_name": "vostankovich",
          "author_url": "",
          "post_date": "09/08/2020 11:02:33",
          "content": "<p>Confidence scores could be anything, because they are used only for sorting your predictions. So, it can be larger than 1. It actually could be any data type that could be sorted.</p>",
          "votes": null,
          "replies": []
        }
      ]
    }
  ],
  "raw_markdown_by_id": {
    "996342": "Hello, For those who were struggling to implement GAP metric, they can also use micro average precision score from sklearn. `sklearn.metric.average_precision_score` . This will allow many competitors to track the metrics they wanted to maximize. \nP.S. open to suggestions and correct me if something is wrong.",
    "998607": "Hosts have open-sourced the code to compute GAP. \nhttps://github.com/tensorflow/models/blob/master/research/delf/delf/python/google_landmarks_dataset/compute_recognition_metrics.py",
    "1002633": "What should be the range of GAP?\n\nShould it be between 0 and 1 or it is also possible that it's value is >1?\n\nAsking because for some of the predictions from baseline model are showing values to be >1",
    "1002698": "Confidence scores could be anything, because they are used only for sorting your predictions. So, it can be larger than 1. It actually could be any data type that could be sorted."
  },
  "source": "meta"
}