{
  "id": 120201,
  "title": "Is the bigger the z, the farther the distance",
  "url": "/competitions/pku-autonomous-driving/discussion/120201",
  "author_name": "",
  "post_date": "2019-12-04T10:57:27.019005300Z",
  "votes": 4,
  "comment_count": 6,
  "views": 0,
  "content": "<p>In my understanding, the farther the car goes, the bigger the z.\nBut if you check these two cars in <strong>ID_dfbcb7cbf</strong>\nNo.1 car's xyz gt is [-23, 6.4, 34]\nNo.2 car's xyz gt is [-20, 5.19, 29.8]\nDoes this mean that car 1 is farther than car 2</p>\n\n<p>Obviously, car 1 is closer\nSo what's wrong with my understanding？</p>\n\n<p><img src=\"https://www.googleapis.com/download/storage/v1/b/kaggle-user-content/o/inbox%2F1832323%2F2086f1eb3ad6dd5b74f3b73c3c6125cc%2Fcff207ab94bd8a64dfb3ff201359c95.png?generation=1575457042751376&amp;alt=media\" alt=\"\"></p>",
  "messages": [
    {
      "id": "687429",
      "postDate": "12/04/2019 10:57:27",
      "content": "<p>In my understanding, the farther the car goes, the bigger the z.\nBut if you check these two cars in <strong>ID_dfbcb7cbf</strong>\nNo.1 car's xyz gt is [-23, 6.4, 34]\nNo.2 car's xyz gt is [-20, 5.19, 29.8]\nDoes this mean that car 1 is farther than car 2</p>\n\n<p>Obviously, car 1 is closer\nSo what's wrong with my understanding？</p>\n\n<p><img src=\"https://www.googleapis.com/download/storage/v1/b/kaggle-user-content/o/inbox%2F1832323%2F2086f1eb3ad6dd5b74f3b73c3c6125cc%2Fcff207ab94bd8a64dfb3ff201359c95.png?generation=1575457042751376&amp;alt=media\" alt=\"\"></p>",
      "rawMarkdown": "In my understanding, the farther the car goes, the bigger the z.\nBut if you check these two cars in **ID_dfbcb7cbf**\nNo.1 car's xyz gt is [-23, 6.4, 34]\nNo.2 car's xyz gt is [-20, 5.19, 29.8]\nDoes this mean that car 1 is farther than car 2\n\nObviously, car 1 is closer\nSo what's wrong with my understanding？\n\n![](https://www.googleapis.com/download/storage/v1/b/kaggle-user-content/o/inbox%2F1832323%2F2086f1eb3ad6dd5b74f3b73c3c6125cc%2Fcff207ab94bd8a64dfb3ff201359c95.png?generation=1575457042751376&amp;alt=media)",
      "votes": null
    },
    {
      "id": "687469",
      "postDate": "12/04/2019 11:47:14",
      "content": "<p>i alse check out <a href=\"https://www.kaggle.com/hocop1/centernet-baseline\">CenterNet-baseline</a> for \"from the sky\"\n<img src=\"https://www.googleapis.com/download/storage/v1/b/kaggle-user-content/o/inbox%2F1832323%2F644c3d585413a5b1afcb8ddf5f65fb19%2F2cd2844d8be419202b17fb33d963065.png?generation=1575459951270519&amp;alt=media\" alt=\"\"></p>\n\n<p>green dot is car 2. So there seems to be a little bit of a problem with the dataset.\nIs there anyone with similar findings</p>",
      "rawMarkdown": "i alse check out [CenterNet-baseline](https://www.kaggle.com/hocop1/centernet-baseline) for \"from the sky\"\n![](https://www.googleapis.com/download/storage/v1/b/kaggle-user-content/o/inbox%2F1832323%2F644c3d585413a5b1afcb8ddf5f65fb19%2F2cd2844d8be419202b17fb33d963065.png?generation=1575459951270519&amp;alt=media)\n\ngreen dot is car 2. So there seems to be a little bit of a problem with the dataset.\nIs there anyone with similar findings",
      "votes": null
    },
    {
      "id": "687694",
      "postDate": "12/04/2019 17:08:54",
      "content": "<p>There definetly are some wrong annotations. But GT was made by human (at least i hope so), so errors will always be there. There are probably also some errors in test GT ... you will just have to work with it</p>",
      "rawMarkdown": "There definetly are some wrong annotations. But GT was made by human (at least i hope so), so errors will always be there. There are probably also some errors in test GT ... you will just have to work with it",
      "votes": null
    },
    {
      "id": "687976",
      "postDate": "12/05/2019 04:46:24",
      "content": "<p>Yeah,  i know what you mean,mistake always happen.\nI'm worried about this because the some wrong annotations make generating incorrect mask😐 </p>",
      "rawMarkdown": "Yeah,  i know what you mean,mistake always happen.\nI'm worried about this because the some wrong annotations make generating incorrect mask😐",
      "votes": null
    },
    {
      "id": "688009",
      "postDate": "12/05/2019 06:02:12",
      "content": "<p>This is true in every image recognition task.\nIn the end, still the best model will win if those mistakes are rare -- and they usually are</p>",
      "rawMarkdown": "This is true in every image recognition task.\nIn the end, still the best model will win if those mistakes are rare -- and they usually are",
      "votes": null
    },
    {
      "id": "688074",
      "postDate": "12/05/2019 07:52:14",
      "content": "<p>Thank you for your advice.Learn More!</p>",
      "rawMarkdown": "Thank you for your advice.Learn More!",
      "votes": null
    },
    {
      "id": "688856",
      "postDate": "12/06/2019 07:06:05",
      "content": "<p>So far as I noticed, the ground plane is not the horizontal one, think about that the camera is actually placed at z =50m and looking down with an small angle, this explain exactly the situation.</p>",
      "rawMarkdown": "So far as I noticed, the ground plane is not the horizontal one, think about that the camera is actually placed at z =50m and looking down with an small angle, this explain exactly the situation.",
      "votes": null
    }
  ],
  "comments": [
    {
      "id": 687469,
      "author_name": "frkaka",
      "author_url": "",
      "post_date": "12/04/2019 11:47:14",
      "content": "<p>i alse check out <a href=\"https://www.kaggle.com/hocop1/centernet-baseline\">CenterNet-baseline</a> for \"from the sky\"\n<img src=\"https://www.googleapis.com/download/storage/v1/b/kaggle-user-content/o/inbox%2F1832323%2F644c3d585413a5b1afcb8ddf5f65fb19%2F2cd2844d8be419202b17fb33d963065.png?generation=1575459951270519&amp;alt=media\" alt=\"\"></p>\n\n<p>green dot is car 2. So there seems to be a little bit of a problem with the dataset.\nIs there anyone with similar findings</p>",
      "votes": null,
      "replies": []
    },
    {
      "id": 687694,
      "author_name": "ilu000",
      "author_url": "",
      "post_date": "12/04/2019 17:08:54",
      "content": "<p>There definetly are some wrong annotations. But GT was made by human (at least i hope so), so errors will always be there. There are probably also some errors in test GT ... you will just have to work with it</p>",
      "votes": null,
      "replies": [
        {
          "id": 687976,
          "author_name": "frkaka",
          "author_url": "",
          "post_date": "12/05/2019 04:46:24",
          "content": "<p>Yeah,  i know what you mean,mistake always happen.\nI'm worried about this because the some wrong annotations make generating incorrect mask😐 </p>",
          "votes": null,
          "replies": []
        },
        {
          "id": 688009,
          "author_name": "ilu000",
          "author_url": "",
          "post_date": "12/05/2019 06:02:12",
          "content": "<p>This is true in every image recognition task.\nIn the end, still the best model will win if those mistakes are rare -- and they usually are</p>",
          "votes": null,
          "replies": []
        },
        {
          "id": 688074,
          "author_name": "frkaka",
          "author_url": "",
          "post_date": "12/05/2019 07:52:14",
          "content": "<p>Thank you for your advice.Learn More!</p>",
          "votes": null,
          "replies": []
        }
      ]
    },
    {
      "id": 688856,
      "author_name": "zhibolin",
      "author_url": "",
      "post_date": "12/06/2019 07:06:05",
      "content": "<p>So far as I noticed, the ground plane is not the horizontal one, think about that the camera is actually placed at z =50m and looking down with an small angle, this explain exactly the situation.</p>",
      "votes": null,
      "replies": []
    }
  ],
  "raw_markdown_by_id": {
    "687429": "In my understanding, the farther the car goes, the bigger the z.\nBut if you check these two cars in **ID_dfbcb7cbf**\nNo.1 car's xyz gt is [-23, 6.4, 34]\nNo.2 car's xyz gt is [-20, 5.19, 29.8]\nDoes this mean that car 1 is farther than car 2\n\nObviously, car 1 is closer\nSo what's wrong with my understanding？\n\n![](https://www.googleapis.com/download/storage/v1/b/kaggle-user-content/o/inbox%2F1832323%2F2086f1eb3ad6dd5b74f3b73c3c6125cc%2Fcff207ab94bd8a64dfb3ff201359c95.png?generation=1575457042751376&amp;alt=media)",
    "687469": "i alse check out [CenterNet-baseline](https://www.kaggle.com/hocop1/centernet-baseline) for \"from the sky\"\n![](https://www.googleapis.com/download/storage/v1/b/kaggle-user-content/o/inbox%2F1832323%2F644c3d585413a5b1afcb8ddf5f65fb19%2F2cd2844d8be419202b17fb33d963065.png?generation=1575459951270519&amp;alt=media)\n\ngreen dot is car 2. So there seems to be a little bit of a problem with the dataset.\nIs there anyone with similar findings",
    "687694": "There definetly are some wrong annotations. But GT was made by human (at least i hope so), so errors will always be there. There are probably also some errors in test GT ... you will just have to work with it",
    "687976": "Yeah,  i know what you mean,mistake always happen.\nI'm worried about this because the some wrong annotations make generating incorrect mask😐",
    "688009": "This is true in every image recognition task.\nIn the end, still the best model will win if those mistakes are rare -- and they usually are",
    "688074": "Thank you for your advice.Learn More!",
    "688856": "So far as I noticed, the ground plane is not the horizontal one, think about that the camera is actually placed at z =50m and looking down with an small angle, this explain exactly the situation."
  },
  "source": "meta"
}