{
  "id": 266751,
  "title": "Data Set understanding",
  "url": "/competitions/landmark-retrieval-2021/discussion/266751",
  "author_name": "",
  "post_date": "2021-08-20T09:10:29.714361500Z",
  "votes": 5,
  "comment_count": 7,
  "views": 0,
  "content": "<p>1) Train Directory is image directory which use to train model.<br>\n2) Test Directory for your testing model.<br>\n3) What is for Index Directory ,what is relation and how to use it ?</p>",
  "messages": [
    {
      "id": "1482773",
      "postDate": "08/20/2021 09:10:29",
      "content": "<p>1) Train Directory is image directory which use to train model.<br>\n2) Test Directory for your testing model.<br>\n3) What is for Index Directory ,what is relation and how to use it ?</p>",
      "rawMarkdown": "1) Train Directory is image directory which use to train model.\n2) Test Directory for your testing model.\n3) What is for Index Directory ,what is relation and how to use it ?",
      "votes": null
    },
    {
      "id": "1488838",
      "postDate": "08/24/2021 14:24:34",
      "content": "<p>Test Images are your query images, which mean you'll use them to pick similar images in the index folder.</p>\n<p>Suppose you have test image ID: 00084cdf8f600d00.jpg<br>\nYou will then find all images in the index folder which are similar to 00084cdf8f600d00.jpg</p>",
      "rawMarkdown": "Test Images are your query images, which mean you'll use them to pick similar images in the index folder.\n\nSuppose you have test image ID: 00084cdf8f600d00.jpg\nYou will then find all images in the index folder which are similar to 00084cdf8f600d00.jpg",
      "votes": null
    },
    {
      "id": "1489567",
      "postDate": "08/25/2021 05:16:47",
      "content": "<p>Thanks for comment and make me understand the dataset.</p>",
      "rawMarkdown": "Thanks for comment and make me understand the dataset.",
      "votes": null
    },
    {
      "id": "1489955",
      "postDate": "08/25/2021 12:02:13",
      "content": "<p>is there any hint on how much larger test and/or index will be in the submission run? I tried to find some clues, but couldn't find any.</p>",
      "rawMarkdown": "is there any hint on how much larger test and/or index will be in the submission run? I tried to find some clues, but couldn't find any.",
      "votes": null
    },
    {
      "id": "1490075",
      "postDate": "08/25/2021 13:34:34",
      "content": "<p>No hint or idea from dataset l☹️☹️|</p>",
      "rawMarkdown": "No hint or idea from dataset l☹️☹️|",
      "votes": null
    },
    {
      "id": "1490724",
      "postDate": "08/25/2021 20:06:34",
      "content": "<p>I still don't understand the test set situation. I understand that I should use images from the test directory to query (find similar images) images in the index directory. But is there any ground truth information? Something like train.csv file but for test? How can I verify how good am I doing on the test set?</p>",
      "rawMarkdown": "I still don't understand the test set situation. I understand that I should use images from the test directory to query (find similar images) images in the index directory. But is there any ground truth information? Something like train.csv file but for test? How can I verify how good am I doing on the test set?",
      "votes": null
    },
    {
      "id": "1491295",
      "postDate": "08/26/2021 09:33:02",
      "content": "<p>same problem. Can't identify the what logic required ?</p>",
      "rawMarkdown": "same problem. Can't identify the what logic required ?",
      "votes": null
    },
    {
      "id": "1491527",
      "postDate": "08/26/2021 12:56:17",
      "content": "<p>Maybe you could understand the situation like this:<br>\nYou'll first train your model to answer the question: \"Which lankmark_id should I give to this image?\"<br>\nThen you will use your model to predict on the test set by answering the question above<br>\nNext you will give a prediction on the index set and answering question again (but be careful with junk images!)<br>\nRemember that test's ground truths are their id + matched index images </p>",
      "rawMarkdown": "Maybe you could understand the situation like this:\nYou'll first train your model to answer the question: \"Which lankmark_id should I give to this image?\"\nThen you will use your model to predict on the test set by answering the question above\nNext you will give a prediction on the index set and answering question again (but be careful with junk images!)\nRemember that test's ground truths are their id + matched index images",
      "votes": null
    }
  ],
  "comments": [
    {
      "id": 1488838,
      "author_name": "tongkhangte",
      "author_url": "",
      "post_date": "08/24/2021 14:24:34",
      "content": "<p>Test Images are your query images, which mean you'll use them to pick similar images in the index folder.</p>\n<p>Suppose you have test image ID: 00084cdf8f600d00.jpg<br>\nYou will then find all images in the index folder which are similar to 00084cdf8f600d00.jpg</p>",
      "votes": null,
      "replies": [
        {
          "id": 1489567,
          "author_name": "ankurgupta29",
          "author_url": "",
          "post_date": "08/25/2021 05:16:47",
          "content": "<p>Thanks for comment and make me understand the dataset.</p>",
          "votes": null,
          "replies": []
        },
        {
          "id": 1489955,
          "author_name": "ilu000",
          "author_url": "",
          "post_date": "08/25/2021 12:02:13",
          "content": "<p>is there any hint on how much larger test and/or index will be in the submission run? I tried to find some clues, but couldn't find any.</p>",
          "votes": null,
          "replies": []
        },
        {
          "id": 1490075,
          "author_name": "ankurgupta29",
          "author_url": "",
          "post_date": "08/25/2021 13:34:34",
          "content": "<p>No hint or idea from dataset l☹️☹️|</p>",
          "votes": null,
          "replies": []
        }
      ]
    },
    {
      "id": 1490724,
      "author_name": "mrekmrik",
      "author_url": "",
      "post_date": "08/25/2021 20:06:34",
      "content": "<p>I still don't understand the test set situation. I understand that I should use images from the test directory to query (find similar images) images in the index directory. But is there any ground truth information? Something like train.csv file but for test? How can I verify how good am I doing on the test set?</p>",
      "votes": null,
      "replies": [
        {
          "id": 1491295,
          "author_name": "ankurgupta29",
          "author_url": "",
          "post_date": "08/26/2021 09:33:02",
          "content": "<p>same problem. Can't identify the what logic required ?</p>",
          "votes": null,
          "replies": []
        },
        {
          "id": 1491527,
          "author_name": "tongkhangte",
          "author_url": "",
          "post_date": "08/26/2021 12:56:17",
          "content": "<p>Maybe you could understand the situation like this:<br>\nYou'll first train your model to answer the question: \"Which lankmark_id should I give to this image?\"<br>\nThen you will use your model to predict on the test set by answering the question above<br>\nNext you will give a prediction on the index set and answering question again (but be careful with junk images!)<br>\nRemember that test's ground truths are their id + matched index images </p>",
          "votes": null,
          "replies": []
        }
      ]
    }
  ],
  "raw_markdown_by_id": {
    "1482773": "1) Train Directory is image directory which use to train model.\n2) Test Directory for your testing model.\n3) What is for Index Directory ,what is relation and how to use it ?",
    "1488838": "Test Images are your query images, which mean you'll use them to pick similar images in the index folder.\n\nSuppose you have test image ID: 00084cdf8f600d00.jpg\nYou will then find all images in the index folder which are similar to 00084cdf8f600d00.jpg",
    "1489567": "Thanks for comment and make me understand the dataset.",
    "1489955": "is there any hint on how much larger test and/or index will be in the submission run? I tried to find some clues, but couldn't find any.",
    "1490075": "No hint or idea from dataset l☹️☹️|",
    "1490724": "I still don't understand the test set situation. I understand that I should use images from the test directory to query (find similar images) images in the index directory. But is there any ground truth information? Something like train.csv file but for test? How can I verify how good am I doing on the test set?",
    "1491295": "same problem. Can't identify the what logic required ?",
    "1491527": "Maybe you could understand the situation like this:\nYou'll first train your model to answer the question: \"Which lankmark_id should I give to this image?\"\nThen you will use your model to predict on the test set by answering the question above\nNext you will give a prediction on the index set and answering question again (but be careful with junk images!)\nRemember that test's ground truths are their id + matched index images"
  },
  "source": "meta"
}