{
  "id": 172838,
  "title": "Getting started with landmark recognition?",
  "url": "/competitions/landmark-recognition-2020/discussion/172838",
  "author_name": "",
  "post_date": "2020-08-06T17:10:03.463186800Z",
  "votes": 7,
  "comment_count": 6,
  "views": 0,
  "content": "<p>In addition to this fine <a href=\"https://www.kaggle.com/c/landmark-recognition-2020/discussion/170989\"><strong>collection</strong></a> of past competition solutions, what resources do you recommend to getting started with landmark recognition (bonus point in Pytorch)? Thanks in advance for the links!</p>",
  "messages": [
    {
      "id": "960783",
      "postDate": "08/06/2020 17:10:03",
      "content": "<p>In addition to this fine <a href=\"https://www.kaggle.com/c/landmark-recognition-2020/discussion/170989\"><strong>collection</strong></a> of past competition solutions, what resources do you recommend to getting started with landmark recognition (bonus point in Pytorch)? Thanks in advance for the links!</p>",
      "rawMarkdown": "In addition to this fine [**collection**](https://www.kaggle.com/c/landmark-recognition-2020/discussion/170989) of past competition solutions, what resources do you recommend to getting started with landmark recognition (bonus point in Pytorch)? Thanks in advance for the links!",
      "votes": null
    },
    {
      "id": "961630",
      "postDate": "08/07/2020 10:59:13",
      "content": "<p>How do handle such large data ? Downloading Locally or via small resolution data ?</p>",
      "rawMarkdown": "How do handle such large data ? Downloading Locally or via small resolution data ?",
      "votes": null
    },
    {
      "id": "979335",
      "postDate": "08/20/2020 19:08:31",
      "content": "<p>I think <a href=\"https://arxiv.org/abs/1503.03832\" target=\"_blank\">https://arxiv.org/abs/1503.03832</a> is relevant enough. since face recognition is a kind of instance recognition and the same method could be applied.</p>\n<p>That was what I was going for anyways. Do you have any other insights?</p>",
      "rawMarkdown": "I think https://arxiv.org/abs/1503.03832 is relevant enough. since face recognition is a kind of instance recognition and the same method could be applied.\n\nThat was what I was going for anyways. Do you have any other insights?",
      "votes": null
    },
    {
      "id": "981585",
      "postDate": "08/22/2020 15:18:21",
      "content": "<p>Thanks for the reply. I will check the reference and let you know. </p>",
      "rawMarkdown": "Thanks for the reply. I will check the reference and let you know.",
      "votes": null
    },
    {
      "id": "981587",
      "postDate": "08/22/2020 15:20:12",
      "content": "<p>Well, there is a possibility to download everything and resize it. You can also look for resized datasets here in Kaggle. I will add a link if I find one here. </p>",
      "rawMarkdown": "Well, there is a possibility to download everything and resize it. You can also look for resized datasets here in Kaggle. I will add a link if I find one here.",
      "votes": null
    },
    {
      "id": "981623",
      "postDate": "08/22/2020 15:36:52",
      "content": "<p>Never mind that. The time it takes to train such a model is HUGE. Besides, it doesn't implement any local feature extraction pipeline which would give poor results.</p>\n<p>You should see into DELG though.</p>",
      "rawMarkdown": "Never mind that. The time it takes to train such a model is HUGE. Besides, it doesn't implement any local feature extraction pipeline which would give poor results.\n\nYou should see into DELG though.",
      "votes": null
    },
    {
      "id": "1000563",
      "postDate": "09/06/2020 16:09:47",
      "content": "<p>Yes, it seems that using a pre-trained DELG model is a good starting point. That's the approach I am pursuing. I am working on a notebook exploring this topic but it is far from being ready. :D </p>",
      "rawMarkdown": "Yes, it seems that using a pre-trained DELG model is a good starting point. That's the approach I am pursuing. I am working on a notebook exploring this topic but it is far from being ready. :D",
      "votes": null
    }
  ],
  "comments": [
    {
      "id": 961630,
      "author_name": "swarajshinde",
      "author_url": "",
      "post_date": "08/07/2020 10:59:13",
      "content": "<p>How do handle such large data ? Downloading Locally or via small resolution data ?</p>",
      "votes": null,
      "replies": [
        {
          "id": 981587,
          "author_name": "yassinealouini",
          "author_url": "",
          "post_date": "08/22/2020 15:20:12",
          "content": "<p>Well, there is a possibility to download everything and resize it. You can also look for resized datasets here in Kaggle. I will add a link if I find one here. </p>",
          "votes": null,
          "replies": []
        }
      ]
    },
    {
      "id": 979335,
      "author_name": "darkcube",
      "author_url": "",
      "post_date": "08/20/2020 19:08:31",
      "content": "<p>I think <a href=\"https://arxiv.org/abs/1503.03832\" target=\"_blank\">https://arxiv.org/abs/1503.03832</a> is relevant enough. since face recognition is a kind of instance recognition and the same method could be applied.</p>\n<p>That was what I was going for anyways. Do you have any other insights?</p>",
      "votes": null,
      "replies": [
        {
          "id": 981585,
          "author_name": "yassinealouini",
          "author_url": "",
          "post_date": "08/22/2020 15:18:21",
          "content": "<p>Thanks for the reply. I will check the reference and let you know. </p>",
          "votes": null,
          "replies": []
        },
        {
          "id": 981623,
          "author_name": "darkcube",
          "author_url": "",
          "post_date": "08/22/2020 15:36:52",
          "content": "<p>Never mind that. The time it takes to train such a model is HUGE. Besides, it doesn't implement any local feature extraction pipeline which would give poor results.</p>\n<p>You should see into DELG though.</p>",
          "votes": null,
          "replies": []
        },
        {
          "id": 1000563,
          "author_name": "yassinealouini",
          "author_url": "",
          "post_date": "09/06/2020 16:09:47",
          "content": "<p>Yes, it seems that using a pre-trained DELG model is a good starting point. That's the approach I am pursuing. I am working on a notebook exploring this topic but it is far from being ready. :D </p>",
          "votes": null,
          "replies": []
        }
      ]
    }
  ],
  "raw_markdown_by_id": {
    "960783": "In addition to this fine [**collection**](https://www.kaggle.com/c/landmark-recognition-2020/discussion/170989) of past competition solutions, what resources do you recommend to getting started with landmark recognition (bonus point in Pytorch)? Thanks in advance for the links!",
    "961630": "How do handle such large data ? Downloading Locally or via small resolution data ?",
    "979335": "I think https://arxiv.org/abs/1503.03832 is relevant enough. since face recognition is a kind of instance recognition and the same method could be applied.\n\nThat was what I was going for anyways. Do you have any other insights?",
    "981585": "Thanks for the reply. I will check the reference and let you know.",
    "981587": "Well, there is a possibility to download everything and resize it. You can also look for resized datasets here in Kaggle. I will add a link if I find one here.",
    "981623": "Never mind that. The time it takes to train such a model is HUGE. Besides, it doesn't implement any local feature extraction pipeline which would give poor results.\n\nYou should see into DELG though.",
    "1000563": "Yes, it seems that using a pre-trained DELG model is a good starting point. That's the approach I am pursuing. I am working on a notebook exploring this topic but it is far from being ready. :D"
  },
  "source": "meta"
}