{
  "id": 318222,
  "title": "Duplication in test dataset",
  "url": "/competitions/snakeclef2022/discussion/318222",
  "author_name": "",
  "post_date": "2022-04-11T07:52:40.586819600Z",
  "votes": null,
  "comment_count": 3,
  "views": 0,
  "content": "<p>Hi~ I have found that there are many cases in the test meta data (.csv file) that one observationId corresponding to multiple images. How should we deal with these cases? In my understanding this means the same object and should be tagged with only one class label per observationId, so we should remove the duplications in our submission files. Is it right? Thanks~</p>",
  "messages": [
    {
      "id": "1751910",
      "postDate": "04/11/2022 07:52:40",
      "content": "<p>Hi~ I have found that there are many cases in the test meta data (.csv file) that one observationId corresponding to multiple images. How should we deal with these cases? In my understanding this means the same object and should be tagged with only one class label per observationId, so we should remove the duplications in our submission files. Is it right? Thanks~</p>",
      "rawMarkdown": "Hi~ I have found that there are many cases in the test meta data (.csv file) that one observationId corresponding to multiple images. How should we deal with these cases? In my understanding this means the same object and should be tagged with only one class label per observationId, so we should remove the duplications in our submission files. Is it right? Thanks~",
      "votes": null
    },
    {
      "id": "1751916",
      "postDate": "04/11/2022 08:00:58",
      "content": "<p>Hi <a href=\"https://www.kaggle.com/edogawachia\" target=\"_blank\">@edogawachia</a>,</p>\n<p>That is the wrong assumption. There should not be any duplicated images. Observation is composed of different images of the same individual.</p>\n<p>Thus, you should \"somehow\" prepare a prediction from all the images. </p>\n<p>Hope it's clear now :)</p>\n<p>Best,<br>\nLukas</p>",
      "rawMarkdown": "Hi @edogawachia,\n\nThat is the wrong assumption. There should not be any duplicated images. Observation is composed of different images of the same individual.\n\nThus, you should \"somehow\" prepare a prediction from all the images. \n\nHope it's clear now :)\n\nBest,\nLukas",
      "votes": null
    },
    {
      "id": "1751947",
      "postDate": "04/11/2022 08:31:50",
      "content": "<p>Got it! Thanks for your quick reply~ 👍</p>",
      "rawMarkdown": "Got it! Thanks for your quick reply~ 👍",
      "votes": null
    },
    {
      "id": "1814989",
      "postDate": "06/08/2022 14:40:24",
      "content": "<p>Please refer to my topic in discussion, I need your help, your upvote, thanks. <a href=\"https://www.kaggle.com/edogawachia\" target=\"_blank\">@edogawachia</a> </p>",
      "rawMarkdown": "Please refer to my topic in discussion, I need your help, your upvote, thanks. @edogawachia",
      "votes": null
    }
  ],
  "comments": [
    {
      "id": 1751916,
      "author_name": "picekl",
      "author_url": "",
      "post_date": "04/11/2022 08:00:58",
      "content": "<p>Hi <a href=\"https://www.kaggle.com/edogawachia\" target=\"_blank\">@edogawachia</a>,</p>\n<p>That is the wrong assumption. There should not be any duplicated images. Observation is composed of different images of the same individual.</p>\n<p>Thus, you should \"somehow\" prepare a prediction from all the images. </p>\n<p>Hope it's clear now :)</p>\n<p>Best,<br>\nLukas</p>",
      "votes": null,
      "replies": [
        {
          "id": 1751947,
          "author_name": "edogawachia",
          "author_url": "",
          "post_date": "04/11/2022 08:31:50",
          "content": "<p>Got it! Thanks for your quick reply~ 👍</p>",
          "votes": null,
          "replies": []
        }
      ]
    },
    {
      "id": 1814989,
      "author_name": "abuduori",
      "author_url": "",
      "post_date": "06/08/2022 14:40:24",
      "content": "<p>Please refer to my topic in discussion, I need your help, your upvote, thanks. <a href=\"https://www.kaggle.com/edogawachia\" target=\"_blank\">@edogawachia</a> </p>",
      "votes": null,
      "replies": []
    }
  ],
  "raw_markdown_by_id": {
    "1751910": "Hi~ I have found that there are many cases in the test meta data (.csv file) that one observationId corresponding to multiple images. How should we deal with these cases? In my understanding this means the same object and should be tagged with only one class label per observationId, so we should remove the duplications in our submission files. Is it right? Thanks~",
    "1751916": "Hi @edogawachia,\n\nThat is the wrong assumption. There should not be any duplicated images. Observation is composed of different images of the same individual.\n\nThus, you should \"somehow\" prepare a prediction from all the images. \n\nHope it's clear now :)\n\nBest,\nLukas",
    "1751947": "Got it! Thanks for your quick reply~ 👍",
    "1814989": "Please refer to my topic in discussion, I need your help, your upvote, thanks. @edogawachia"
  },
  "source": "meta"
}