{
  "id": 176319,
  "title": "How many unique landmarks are in the testing set?",
  "url": "/competitions/landmark-recognition-2020/discussion/176319",
  "author_name": "",
  "post_date": "2020-08-21T09:50:12.439717200Z",
  "votes": 2,
  "comment_count": 4,
  "views": 0,
  "content": "<p>How many unique landmarks are in the test set? &lt;=81313? Are they all present in the training dataset? I mean, is it possible to find samples which landmarks are not in the training set?</p>",
  "messages": [
    {
      "id": "980046",
      "postDate": "08/21/2020 09:50:12",
      "content": "<p>How many unique landmarks are in the test set? &lt;=81313? Are they all present in the training dataset? I mean, is it possible to find samples which landmarks are not in the training set?</p>",
      "rawMarkdown": "How many unique landmarks are in the test set? <=81313? Are they all present in the training dataset? I mean, is it possible to find samples which landmarks are not in the training set?",
      "votes": null
    },
    {
      "id": "980272",
      "postDate": "08/21/2020 12:58:36",
      "content": "<blockquote>\n  <p>This 100k subset contains all of the training set images associated with the landmarks in the private test set</p>\n</blockquote>\n<p>Yes, the landmarks would be a subset of the landmarks present in the training sets. Although, quite a few images in test data do not have any landmark associated with them, and the score is penalised when we predict a non landmark image as a landmark with high confidence</p>",
      "rawMarkdown": "> This 100k subset contains all of the training set images associated with the landmarks in the private test set\n\nYes, the landmarks would be a subset of the landmarks present in the training sets. Although, quite a few images in test data do not have any landmark associated with them, and the score is penalised when we predict a non landmark image as a landmark with high confidence",
      "votes": null
    },
    {
      "id": "983150",
      "postDate": "08/24/2020 04:50:05",
      "content": "<p><a href=\"/aviraljn9\">@aviraljn9</a> Thanks for the explanation. I have a follow-up question: What should be the labels of the images that doesn't have any landmark?</p>",
      "rawMarkdown": "aviraljn9 Thanks for the explanation. I have a follow-up question: What should be the labels of the images that doesn't have any landmark?",
      "votes": null
    },
    {
      "id": "985526",
      "postDate": "08/25/2020 19:30:40",
      "content": "<p><a href=\"https://www.kaggle.com/azaemon\" target=\"_blank\">@azaemon</a> when predicting an image that is no landmark the prediction should be empty ==&gt; ' '.</p>",
      "rawMarkdown": "azaemon when predicting an image that is no landmark the prediction should be empty ==> ' '.",
      "votes": null
    },
    {
      "id": "985551",
      "postDate": "08/25/2020 20:10:10",
      "content": "<p>Thanks for clarifying.</p>",
      "rawMarkdown": "Thanks for clarifying.",
      "votes": null
    }
  ],
  "comments": [
    {
      "id": 980272,
      "author_name": "aviraljn9",
      "author_url": "",
      "post_date": "08/21/2020 12:58:36",
      "content": "<blockquote>\n  <p>This 100k subset contains all of the training set images associated with the landmarks in the private test set</p>\n</blockquote>\n<p>Yes, the landmarks would be a subset of the landmarks present in the training sets. Although, quite a few images in test data do not have any landmark associated with them, and the score is penalised when we predict a non landmark image as a landmark with high confidence</p>",
      "votes": null,
      "replies": [
        {
          "id": 983150,
          "author_name": "azaemon",
          "author_url": "",
          "post_date": "08/24/2020 04:50:05",
          "content": "<p><a href=\"/aviraljn9\">@aviraljn9</a> Thanks for the explanation. I have a follow-up question: What should be the labels of the images that doesn't have any landmark?</p>",
          "votes": null,
          "replies": []
        },
        {
          "id": 985526,
          "author_name": "rsmits",
          "author_url": "",
          "post_date": "08/25/2020 19:30:40",
          "content": "<p><a href=\"https://www.kaggle.com/azaemon\" target=\"_blank\">@azaemon</a> when predicting an image that is no landmark the prediction should be empty ==&gt; ' '.</p>",
          "votes": null,
          "replies": []
        },
        {
          "id": 985551,
          "author_name": "azaemon",
          "author_url": "",
          "post_date": "08/25/2020 20:10:10",
          "content": "<p>Thanks for clarifying.</p>",
          "votes": null,
          "replies": []
        }
      ]
    }
  ],
  "raw_markdown_by_id": {
    "980046": "How many unique landmarks are in the test set? <=81313? Are they all present in the training dataset? I mean, is it possible to find samples which landmarks are not in the training set?",
    "980272": "> This 100k subset contains all of the training set images associated with the landmarks in the private test set\n\nYes, the landmarks would be a subset of the landmarks present in the training sets. Although, quite a few images in test data do not have any landmark associated with them, and the score is penalised when we predict a non landmark image as a landmark with high confidence",
    "983150": "aviraljn9 Thanks for the explanation. I have a follow-up question: What should be the labels of the images that doesn't have any landmark?",
    "985526": "azaemon when predicting an image that is no landmark the prediction should be empty ==> ' '.",
    "985551": "Thanks for clarifying."
  },
  "source": "meta"
}