{
  "id": 266494,
  "title": "Clarify about the number of classes",
  "url": "/competitions/landmark-recognition-2021/discussion/266494",
  "author_name": "",
  "post_date": "2021-08-19T10:30:44.085656600Z",
  "votes": null,
  "comment_count": 2,
  "views": 0,
  "content": "<p>Hello guys, in train set we have 81313 classes. And \"the private training set contains only a 100k subset of the total public training set. This 100k subset contains all of the training set images associated with the landmarks in the private test set.\" So will the private test set have 81313 classes or smaller classes? I wonder if the output of model is exactly 81313 or depend on the private training set better. Thanks all!</p>",
  "messages": [
    {
      "id": "1481126",
      "postDate": "08/19/2021 10:30:44",
      "content": "<p>Hello guys, in train set we have 81313 classes. And \"the private training set contains only a 100k subset of the total public training set. This 100k subset contains all of the training set images associated with the landmarks in the private test set.\" So will the private test set have 81313 classes or smaller classes? I wonder if the output of model is exactly 81313 or depend on the private training set better. Thanks all!</p>",
      "rawMarkdown": "Hello guys, in train set we have 81313 classes. And \"the private training set contains only a 100k subset of the total public training set. This 100k subset contains all of the training set images associated with the landmarks in the private test set.\" So will the private test set have 81313 classes or smaller classes? I wonder if the output of model is exactly 81313 or depend on the private training set better. Thanks all!",
      "votes": null
    },
    {
      "id": "1482842",
      "postDate": "08/20/2021 10:07:13",
      "content": "<p>The test set contains both landmark images and non-landmark images. For each image in the test set you need to either assign a landmark from the 81313 classes. <strong>You need to assign the landmark_id, not the class from your model</strong>.</p>\n<p>The <a href=\"https://www.kaggle.com/c/landmark-recognition-2021/overview/evaluation\" target=\"_blank\">evaluation description</a> gives a clear explanation.</p>\n<pre><code>For each id in the test set, you can predict at most one landmark and its corresponding confidence score. Some query images may contain no landmarks. You may decide not to predict any result for a given query, by submitting an empty prediction.\n</code></pre>",
      "rawMarkdown": "The test set contains both landmark images and non-landmark images. For each image in the test set you need to either assign a landmark from the 81313 classes. **You need to assign the landmark_id, not the class from your model**.\n\nThe [evaluation description](https://www.kaggle.com/c/landmark-recognition-2021/overview/evaluation) gives a clear explanation.\n\n```\nFor each id in the test set, you can predict at most one landmark and its corresponding confidence score. Some query images may contain no landmarks. You may decide not to predict any result for a given query, by submitting an empty prediction.\n```",
      "votes": null
    },
    {
      "id": "1483317",
      "postDate": "08/20/2021 15:16:32",
      "content": "<p>Tks for your reply.</p>",
      "rawMarkdown": "Tks for your reply.",
      "votes": null
    }
  ],
  "comments": [
    {
      "id": 1482842,
      "author_name": "markwijkhuizen",
      "author_url": "",
      "post_date": "08/20/2021 10:07:13",
      "content": "<p>The test set contains both landmark images and non-landmark images. For each image in the test set you need to either assign a landmark from the 81313 classes. <strong>You need to assign the landmark_id, not the class from your model</strong>.</p>\n<p>The <a href=\"https://www.kaggle.com/c/landmark-recognition-2021/overview/evaluation\" target=\"_blank\">evaluation description</a> gives a clear explanation.</p>\n<pre><code>For each id in the test set, you can predict at most one landmark and its corresponding confidence score. Some query images may contain no landmarks. You may decide not to predict any result for a given query, by submitting an empty prediction.\n</code></pre>",
      "votes": null,
      "replies": [
        {
          "id": 1483317,
          "author_name": "",
          "author_url": "",
          "post_date": "08/20/2021 15:16:32",
          "content": "<p>Tks for your reply.</p>",
          "votes": null,
          "replies": []
        }
      ]
    }
  ],
  "raw_markdown_by_id": {
    "1481126": "Hello guys, in train set we have 81313 classes. And \"the private training set contains only a 100k subset of the total public training set. This 100k subset contains all of the training set images associated with the landmarks in the private test set.\" So will the private test set have 81313 classes or smaller classes? I wonder if the output of model is exactly 81313 or depend on the private training set better. Thanks all!",
    "1482842": "The test set contains both landmark images and non-landmark images. For each image in the test set you need to either assign a landmark from the 81313 classes. **You need to assign the landmark_id, not the class from your model**.\n\nThe [evaluation description](https://www.kaggle.com/c/landmark-recognition-2021/overview/evaluation) gives a clear explanation.\n\n```\nFor each id in the test set, you can predict at most one landmark and its corresponding confidence score. Some query images may contain no landmarks. You may decide not to predict any result for a given query, by submitting an empty prediction.\n```",
    "1483317": "Tks for your reply."
  },
  "source": "meta"
}