{
  "id": 117082,
  "title": "Few huge shakeups. Any idea? ",
  "url": "/competitions/3d-object-detection-for-autonomous-vehicles/discussion/117082",
  "author_name": "",
  "post_date": "2019-11-13T09:36:10.076942400Z",
  "votes": 14,
  "comment_count": 7,
  "views": 0,
  "content": "<p>With almost no shakeup, what happened with these three teams? Any idea why?</p>\n\n<p><img src=\"https://www.googleapis.com/download/storage/v1/b/kaggle-user-content/o/inbox%2F1788308%2F75c31a3ea54c227b4dff449b14118854%2FScreenshot_20191113-102642__01.jpg?generation=1573637476614709&amp;alt=media\" alt=\"\"></p>\n\n<p><img src=\"https://www.googleapis.com/download/storage/v1/b/kaggle-user-content/o/inbox%2F1788308%2F346ff4c7cb2c9a62abb4ac0aa032a47f%2FScreenshot_20191113-102635__01.jpg?generation=1573637643077823&amp;alt=media\" alt=\"\"></p>\n\n<p><img src=\"https://www.googleapis.com/download/storage/v1/b/kaggle-user-content/o/inbox%2F1788308%2Fb1a2e5fa82579863ce5793b85a949226%2FScreenshot_20191113-102621__01.jpg?generation=1573637659516373&amp;alt=media\" alt=\"\"></p>",
  "messages": [
    {
      "id": "671893",
      "postDate": "11/13/2019 09:36:10",
      "content": "<p>With almost no shakeup, what happened with these three teams? Any idea why?</p>\n\n<p><img src=\"https://www.googleapis.com/download/storage/v1/b/kaggle-user-content/o/inbox%2F1788308%2F75c31a3ea54c227b4dff449b14118854%2FScreenshot_20191113-102642__01.jpg?generation=1573637476614709&amp;alt=media\" alt=\"\"></p>\n\n<p><img src=\"https://www.googleapis.com/download/storage/v1/b/kaggle-user-content/o/inbox%2F1788308%2F346ff4c7cb2c9a62abb4ac0aa032a47f%2FScreenshot_20191113-102635__01.jpg?generation=1573637643077823&amp;alt=media\" alt=\"\"></p>\n\n<p><img src=\"https://www.googleapis.com/download/storage/v1/b/kaggle-user-content/o/inbox%2F1788308%2Fb1a2e5fa82579863ce5793b85a949226%2FScreenshot_20191113-102621__01.jpg?generation=1573637659516373&amp;alt=media\" alt=\"\"></p>",
      "rawMarkdown": "With almost no shakeup, what happened with these three teams? Any idea why?\n\n![](https://www.googleapis.com/download/storage/v1/b/kaggle-user-content/o/inbox%2F1788308%2F75c31a3ea54c227b4dff449b14118854%2FScreenshot_20191113-102642__01.jpg?generation=1573637476614709&amp;alt=media)\n\n![](https://www.googleapis.com/download/storage/v1/b/kaggle-user-content/o/inbox%2F1788308%2F346ff4c7cb2c9a62abb4ac0aa032a47f%2FScreenshot_20191113-102635__01.jpg?generation=1573637643077823&amp;alt=media)\n\n\n![](https://www.googleapis.com/download/storage/v1/b/kaggle-user-content/o/inbox%2F1788308%2Fb1a2e5fa82579863ce5793b85a949226%2FScreenshot_20191113-102621__01.jpg?generation=1573637659516373&amp;alt=media)",
      "votes": null
    },
    {
      "id": "671962",
      "postDate": "11/13/2019 11:28:55",
      "content": "<p>Normally, this happens when you completely ignore local validation and validate models on the leaderboard. But these are extreme cases for sure. </p>",
      "rawMarkdown": "Normally, this happens when you completely ignore local validation and validate models on the leaderboard. But these are extreme cases for sure.",
      "votes": null
    },
    {
      "id": "671993",
      "postDate": "11/13/2019 12:37:13",
      "content": "<p>The selection of the final two models plays a big role many times in the final leaderboard!</p>",
      "rawMarkdown": "The selection of the final two models plays a big role many times in the final leaderboard!",
      "votes": null
    },
    {
      "id": "672081",
      "postDate": "11/13/2019 13:53:36",
      "content": "<p>I think shapeups didn't affect the results, because of the huge amount of bounding boxes in each test set. I found about 920k boxes total.</p>",
      "rawMarkdown": "I think shapeups didn't affect the results, because of the huge amount of bounding boxes in each test set. I found about 920k boxes total.",
      "votes": null
    },
    {
      "id": "672215",
      "postDate": "11/13/2019 16:28:20",
      "content": "<p>I am very confused about this as well. For the steel competition I saw many people fall thousands of places to the bottom and getting a score of 0. How does that work?</p>",
      "rawMarkdown": "I am very confused about this as well. For the steel competition I saw many people fall thousands of places to the bottom and getting a score of 0. How does that work?",
      "votes": null
    },
    {
      "id": "672226",
      "postDate": "11/13/2019 16:55:16",
      "content": "<p>Steel was a different case, there many people were using \"quick submissions\" which didn't account for the private test set resulting in zero scores in private LB.</p>",
      "rawMarkdown": "Steel was a different case, there many people were using \"quick submissions\" which didn't account for the private test set resulting in zero scores in private LB.",
      "votes": null
    },
    {
      "id": "672227",
      "postDate": "11/13/2019 16:59:16",
      "content": "<p>They haven't participated in any other competitions, so I think they are someone's accounts created for this competition entirely, possibly for LB probing. I think they have picked zero score submissions, there's no other explanation.</p>",
      "rawMarkdown": "They haven't participated in any other competitions, so I think they are someone's accounts created for this competition entirely, possibly for LB probing. I think they have picked zero score submissions, there's no other explanation.",
      "votes": null
    },
    {
      "id": "673673",
      "postDate": "11/15/2019 10:06:00",
      "content": "<p>sekiro111 has been removed</p>",
      "rawMarkdown": "sekiro111 has been removed",
      "votes": null
    }
  ],
  "comments": [
    {
      "id": 671962,
      "author_name": "artyomp",
      "author_url": "",
      "post_date": "11/13/2019 11:28:55",
      "content": "<p>Normally, this happens when you completely ignore local validation and validate models on the leaderboard. But these are extreme cases for sure. </p>",
      "votes": null,
      "replies": []
    },
    {
      "id": 671993,
      "author_name": "tayorm",
      "author_url": "",
      "post_date": "11/13/2019 12:37:13",
      "content": "<p>The selection of the final two models plays a big role many times in the final leaderboard!</p>",
      "votes": null,
      "replies": []
    },
    {
      "id": 672081,
      "author_name": "doncalculator",
      "author_url": "",
      "post_date": "11/13/2019 13:53:36",
      "content": "<p>I think shapeups didn't affect the results, because of the huge amount of bounding boxes in each test set. I found about 920k boxes total.</p>",
      "votes": null,
      "replies": []
    },
    {
      "id": 672215,
      "author_name": "cweed28",
      "author_url": "",
      "post_date": "11/13/2019 16:28:20",
      "content": "<p>I am very confused about this as well. For the steel competition I saw many people fall thousands of places to the bottom and getting a score of 0. How does that work?</p>",
      "votes": null,
      "replies": [
        {
          "id": 672226,
          "author_name": "rishabhiitbhu",
          "author_url": "",
          "post_date": "11/13/2019 16:55:16",
          "content": "<p>Steel was a different case, there many people were using \"quick submissions\" which didn't account for the private test set resulting in zero scores in private LB.</p>",
          "votes": null,
          "replies": []
        }
      ]
    },
    {
      "id": 672227,
      "author_name": "artyomp",
      "author_url": "",
      "post_date": "11/13/2019 16:59:16",
      "content": "<p>They haven't participated in any other competitions, so I think they are someone's accounts created for this competition entirely, possibly for LB probing. I think they have picked zero score submissions, there's no other explanation.</p>",
      "votes": null,
      "replies": []
    },
    {
      "id": 673673,
      "author_name": "artyomp",
      "author_url": "",
      "post_date": "11/15/2019 10:06:00",
      "content": "<p>sekiro111 has been removed</p>",
      "votes": null,
      "replies": []
    }
  ],
  "raw_markdown_by_id": {
    "671893": "With almost no shakeup, what happened with these three teams? Any idea why?\n\n![](https://www.googleapis.com/download/storage/v1/b/kaggle-user-content/o/inbox%2F1788308%2F75c31a3ea54c227b4dff449b14118854%2FScreenshot_20191113-102642__01.jpg?generation=1573637476614709&amp;alt=media)\n\n![](https://www.googleapis.com/download/storage/v1/b/kaggle-user-content/o/inbox%2F1788308%2F346ff4c7cb2c9a62abb4ac0aa032a47f%2FScreenshot_20191113-102635__01.jpg?generation=1573637643077823&amp;alt=media)\n\n\n![](https://www.googleapis.com/download/storage/v1/b/kaggle-user-content/o/inbox%2F1788308%2Fb1a2e5fa82579863ce5793b85a949226%2FScreenshot_20191113-102621__01.jpg?generation=1573637659516373&amp;alt=media)",
    "671962": "Normally, this happens when you completely ignore local validation and validate models on the leaderboard. But these are extreme cases for sure.",
    "671993": "The selection of the final two models plays a big role many times in the final leaderboard!",
    "672081": "I think shapeups didn't affect the results, because of the huge amount of bounding boxes in each test set. I found about 920k boxes total.",
    "672215": "I am very confused about this as well. For the steel competition I saw many people fall thousands of places to the bottom and getting a score of 0. How does that work?",
    "672226": "Steel was a different case, there many people were using \"quick submissions\" which didn't account for the private test set resulting in zero scores in private LB.",
    "672227": "They haven't participated in any other competitions, so I think they are someone's accounts created for this competition entirely, possibly for LB probing. I think they have picked zero score submissions, there's no other explanation.",
    "673673": "sekiro111 has been removed"
  },
  "source": "meta"
}