{
  "id": 166335,
  "title": "Query regarding dataset",
  "url": "/competitions/landmark-retrieval-2020/discussion/166335",
  "author_name": "",
  "post_date": "2020-07-12T14:35:47.769109800Z",
  "votes": null,
  "comment_count": 1,
  "views": 0,
  "content": "<p>Can someone clarify why the number of train images are 1.6M and 203k labels instead of 81K ? if the dataset is the GLDv2-train-clean the labels should be 81k according to the paper.</p>",
  "messages": [
    {
      "id": "926156",
      "postDate": "07/12/2020 14:35:47",
      "content": "<p>Can someone clarify why the number of train images are 1.6M and 203k labels instead of 81K ? if the dataset is the GLDv2-train-clean the labels should be 81k according to the paper.</p>",
      "rawMarkdown": "Can someone clarify why the number of train images are 1.6M and 203k labels instead of 81K ? if the dataset is the GLDv2-train-clean the labels should be 81k according to the paper.",
      "votes": null
    },
    {
      "id": "939154",
      "postDate": "07/22/2020 04:11:46",
      "content": "<p>There are only 81K labels. If you download the train.csv from the competitions data page and look at \ntrain.landmark_id.nunique() ##This code is for python\nyou will see that there are only 81K labels 81313 to be precise. \nHope this helps!!!</p>",
      "rawMarkdown": "There are only 81K labels. If you download the train.csv from the competitions data page and look at \ntrain.landmark_id.nunique() ##This code is for python\nyou will see that there are only 81K labels 81313 to be precise. \nHope this helps!!!",
      "votes": null
    }
  ],
  "comments": [
    {
      "id": 939154,
      "author_name": "chandanverma",
      "author_url": "",
      "post_date": "07/22/2020 04:11:46",
      "content": "<p>There are only 81K labels. If you download the train.csv from the competitions data page and look at \ntrain.landmark_id.nunique() ##This code is for python\nyou will see that there are only 81K labels 81313 to be precise. \nHope this helps!!!</p>",
      "votes": null,
      "replies": []
    }
  ],
  "raw_markdown_by_id": {
    "926156": "Can someone clarify why the number of train images are 1.6M and 203k labels instead of 81K ? if the dataset is the GLDv2-train-clean the labels should be 81k according to the paper.",
    "939154": "There are only 81K labels. If you download the train.csv from the competitions data page and look at \ntrain.landmark_id.nunique() ##This code is for python\nyou will see that there are only 81K labels 81313 to be precise. \nHope this helps!!!"
  },
  "source": "meta"
}