{
  "id": 96829,
  "title": "Absolutely no difference from last year?",
  "url": "/competitions/open-images-2019-object-detection/discussion/96829",
  "author_name": "",
  "post_date": "2019-06-23T15:09:43.787985800Z",
  "votes": 8,
  "comment_count": 5,
  "views": 0,
  "content": "<p>Did you guys notice that the public LB scores of team PFDet for 2018 and 2019 are exactly the same: 0.62882 right now! How is this possible for such a huge dataset and a complex task? I would say this means that there is absolutely no difference between two competitions. Team PFDet only has one submission now and they probably just submitted the file from their top 2 solution last year.</p>",
  "messages": [
    {
      "id": "559157",
      "postDate": "06/23/2019 15:09:43",
      "content": "<p>Did you guys notice that the public LB scores of team PFDet for 2018 and 2019 are exactly the same: 0.62882 right now! How is this possible for such a huge dataset and a complex task? I would say this means that there is absolutely no difference between two competitions. Team PFDet only has one submission now and they probably just submitted the file from their top 2 solution last year.</p>",
      "rawMarkdown": "Did you guys notice that the public LB scores of team PFDet for 2018 and 2019 are exactly the same: 0.62882 right now! How is this possible for such a huge dataset and a complex task? I would say this means that there is absolutely no difference between two competitions. Team PFDet only has one submission now and they probably just submitted the file from their top 2 solution last year.",
      "votes": null
    },
    {
      "id": "559242",
      "postDate": "06/23/2019 17:54:16",
      "content": "<p>Same for other users (see n01z3 for instance)\nI wonder if the organizers at least changed the private LB. It would be good to have some feedback from them...</p>",
      "rawMarkdown": "Same for other users (see n01z3 for instance)\nI wonder if the organizers at least changed the private LB. It would be good to have some feedback from them...",
      "votes": null
    },
    {
      "id": "559363",
      "postDate": "06/24/2019 01:27:08",
      "content": "<p>Are there any differences in categories between v4 and v5?</p>",
      "rawMarkdown": "Are there any differences in categories between v4 and v5?",
      "votes": null
    },
    {
      "id": "563133",
      "postDate": "06/28/2019 00:14:49",
      "content": "<p>The train set is different, the test set is same. </p>",
      "rawMarkdown": "The train set is different, the test set is same.",
      "votes": null
    },
    {
      "id": "564583",
      "postDate": "06/29/2019 16:41:27",
      "content": "<p>I was not sure for that, even though I unzip the test_challenge.zip, and its folder name is challenge2018. Now I'm more sure for that.</p>",
      "rawMarkdown": "I was not sure for that, even though I unzip the test_challenge.zip, and its folder name is challenge2018. Now I'm more sure for that.",
      "votes": null
    },
    {
      "id": "564741",
      "postDate": "06/29/2019 22:29:51",
      "content": "<p>For google, this is actually a continuation competition. They do not put too much effort into it (hey, there is not even a welcome thread :) )\nthe training set and labeling has changed a little, some class names have been changed to resolve ambiguous names (like \"horn (musical instrument)\").\nall in all, it is the same competitions. However, a year has passed, so new algorithms, methods etc exist.\nIt is still remarkably difficult to beat PFDet 5m$ computer....</p>",
      "rawMarkdown": "For google, this is actually a continuation competition. They do not put too much effort into it (hey, there is not even a welcome thread :) )\nthe training set and labeling has changed a little, some class names have been changed to resolve ambiguous names (like \"horn (musical instrument)\").\nall in all, it is the same competitions. However, a year has passed, so new algorithms, methods etc exist.\nIt is still remarkably difficult to beat PFDet 5m$ computer....",
      "votes": null
    }
  ],
  "comments": [
    {
      "id": 559242,
      "author_name": "stecasasso",
      "author_url": "",
      "post_date": "06/23/2019 17:54:16",
      "content": "<p>Same for other users (see n01z3 for instance)\nI wonder if the organizers at least changed the private LB. It would be good to have some feedback from them...</p>",
      "votes": null,
      "replies": []
    },
    {
      "id": 559363,
      "author_name": "interneuron",
      "author_url": "",
      "post_date": "06/24/2019 01:27:08",
      "content": "<p>Are there any differences in categories between v4 and v5?</p>",
      "votes": null,
      "replies": []
    },
    {
      "id": 563133,
      "author_name": "artyomp",
      "author_url": "",
      "post_date": "06/28/2019 00:14:49",
      "content": "<p>The train set is different, the test set is same. </p>",
      "votes": null,
      "replies": [
        {
          "id": 564583,
          "author_name": "jiangjixiang",
          "author_url": "",
          "post_date": "06/29/2019 16:41:27",
          "content": "<p>I was not sure for that, even though I unzip the test_challenge.zip, and its folder name is challenge2018. Now I'm more sure for that.</p>",
          "votes": null,
          "replies": []
        }
      ]
    },
    {
      "id": 564741,
      "author_name": "moshel",
      "author_url": "",
      "post_date": "06/29/2019 22:29:51",
      "content": "<p>For google, this is actually a continuation competition. They do not put too much effort into it (hey, there is not even a welcome thread :) )\nthe training set and labeling has changed a little, some class names have been changed to resolve ambiguous names (like \"horn (musical instrument)\").\nall in all, it is the same competitions. However, a year has passed, so new algorithms, methods etc exist.\nIt is still remarkably difficult to beat PFDet 5m$ computer....</p>",
      "votes": null,
      "replies": []
    }
  ],
  "raw_markdown_by_id": {
    "559157": "Did you guys notice that the public LB scores of team PFDet for 2018 and 2019 are exactly the same: 0.62882 right now! How is this possible for such a huge dataset and a complex task? I would say this means that there is absolutely no difference between two competitions. Team PFDet only has one submission now and they probably just submitted the file from their top 2 solution last year.",
    "559242": "Same for other users (see n01z3 for instance)\nI wonder if the organizers at least changed the private LB. It would be good to have some feedback from them...",
    "559363": "Are there any differences in categories between v4 and v5?",
    "563133": "The train set is different, the test set is same.",
    "564583": "I was not sure for that, even though I unzip the test_challenge.zip, and its folder name is challenge2018. Now I'm more sure for that.",
    "564741": "For google, this is actually a continuation competition. They do not put too much effort into it (hey, there is not even a welcome thread :) )\nthe training set and labeling has changed a little, some class names have been changed to resolve ambiguous names (like \"horn (musical instrument)\").\nall in all, it is the same competitions. However, a year has passed, so new algorithms, methods etc exist.\nIt is still remarkably difficult to beat PFDet 5m$ computer...."
  },
  "source": "meta"
}