{
  "id": 116423,
  "title": "Rotation Evaluation findings",
  "url": "/competitions/pku-autonomous-driving/discussion/116423",
  "author_name": "",
  "post_date": "2019-11-09T00:25:55.119236800Z",
  "votes": 7,
  "comment_count": 2,
  "views": 0,
  "content": "<p>I found it interesting how the roll of the car is either -pi or pi (mostly -pi) which, numerically, is a big difference, but visually its almost exactly the same. So I wondered if this competition would evaluate the predictions numerically or visually, and it seems you get a lower score if you have the wrong sign, which in reality is not off by much. I think this also applies to pitch as well.</p>",
  "messages": [
    {
      "id": "668848",
      "postDate": "11/09/2019 00:25:55",
      "content": "<p>I found it interesting how the roll of the car is either -pi or pi (mostly -pi) which, numerically, is a big difference, but visually its almost exactly the same. So I wondered if this competition would evaluate the predictions numerically or visually, and it seems you get a lower score if you have the wrong sign, which in reality is not off by much. I think this also applies to pitch as well.</p>",
      "rawMarkdown": "I found it interesting how the roll of the car is either -pi or pi (mostly -pi) which, numerically, is a big difference, but visually its almost exactly the same. So I wondered if this competition would evaluate the predictions numerically or visually, and it seems you get a lower score if you have the wrong sign, which in reality is not off by much. I think this also applies to pitch as well.",
      "votes": null
    },
    {
      "id": "671844",
      "postDate": "11/13/2019 08:20:32",
      "content": "<p>Hi, I think the evaluation code transform it to quaternion first. Hence, the pi (-pi) does not make a difference if tranformed into quaternion first and then calcuated the radian difference.</p>",
      "rawMarkdown": "Hi, I think the evaluation code transform it to quaternion first. Hence, the pi (-pi) does not make a difference if tranformed into quaternion first and then calcuated the radian difference.",
      "votes": null
    },
    {
      "id": "672019",
      "postDate": "11/13/2019 13:05:42",
      "content": "<p>I suppose so, when I tested it my score didn't go down that much (1%). Hopefully we get clarification or see the official metric code.</p>",
      "rawMarkdown": "I suppose so, when I tested it my score didn't go down that much (1%). Hopefully we get clarification or see the official metric code.",
      "votes": null
    }
  ],
  "comments": [
    {
      "id": 671844,
      "author_name": "stevenwudi",
      "author_url": "",
      "post_date": "11/13/2019 08:20:32",
      "content": "<p>Hi, I think the evaluation code transform it to quaternion first. Hence, the pi (-pi) does not make a difference if tranformed into quaternion first and then calcuated the radian difference.</p>",
      "votes": null,
      "replies": [
        {
          "id": 672019,
          "author_name": "greatgamedota",
          "author_url": "",
          "post_date": "11/13/2019 13:05:42",
          "content": "<p>I suppose so, when I tested it my score didn't go down that much (1%). Hopefully we get clarification or see the official metric code.</p>",
          "votes": null,
          "replies": []
        }
      ]
    }
  ],
  "raw_markdown_by_id": {
    "668848": "I found it interesting how the roll of the car is either -pi or pi (mostly -pi) which, numerically, is a big difference, but visually its almost exactly the same. So I wondered if this competition would evaluate the predictions numerically or visually, and it seems you get a lower score if you have the wrong sign, which in reality is not off by much. I think this also applies to pitch as well.",
    "671844": "Hi, I think the evaluation code transform it to quaternion first. Hence, the pi (-pi) does not make a difference if tranformed into quaternion first and then calcuated the radian difference.",
    "672019": "I suppose so, when I tested it my score didn't go down that much (1%). Hopefully we get clarification or see the official metric code."
  },
  "source": "meta"
}