{
  "id": 109851,
  "title": "Axis definition of Lyft data",
  "url": "/competitions/3d-object-detection-for-autonomous-vehicles/discussion/109851",
  "author_name": "",
  "post_date": "2019-09-23T02:13:50.623802400Z",
  "votes": 6,
  "comment_count": 3,
  "views": 0,
  "content": "<p>Hi,</p>\n\n<p>Could you please give us a clear axis definition of LiDAR in lyft data? For example, in kitti dataset, its axis definition of LiDAR is \"x-front, y-left, z-up\".</p>\n\n<p>Many thanks.</p>",
  "messages": [
    {
      "id": "631949",
      "postDate": "09/23/2019 02:13:50",
      "content": "<p>Hi,</p>\n\n<p>Could you please give us a clear axis definition of LiDAR in lyft data? For example, in kitti dataset, its axis definition of LiDAR is \"x-front, y-left, z-up\".</p>\n\n<p>Many thanks.</p>",
      "rawMarkdown": "Hi,\n\nCould you please give us a clear axis definition of LiDAR in lyft data? For example, in kitti dataset, its axis definition of LiDAR is \"x-front, y-left, z-up\".\n\nMany thanks.",
      "votes": null
    },
    {
      "id": "632265",
      "postDate": "09/23/2019 12:10:46",
      "content": "<p><img src=\"https://www.googleapis.com/download/storage/v1/b/kaggle-user-content/o/inbox%2F761152%2F8b6b2b756e577017acbb9a24c9337561%2Fpic-selected-190923-1738-51.png?generation=1569240554397179&amp;alt=media\" alt=\"\"></p>\n\n<p>Taken from the nuscenes dataset, as lyft also follows the same format we can say that for \nLIDAR_TOP: x-right, y-front, z-up</p>\n\n<p><a href=\"https://arxiv.org/abs/1903.11027\">Paper</a></p>",
      "rawMarkdown": "![](https://www.googleapis.com/download/storage/v1/b/kaggle-user-content/o/inbox%2F761152%2F8b6b2b756e577017acbb9a24c9337561%2Fpic-selected-190923-1738-51.png?generation=1569240554397179&amp;alt=media)\n\nTaken from the nuscenes dataset, as lyft also follows the same format we can say that for \nLIDAR_TOP: x-right, y-front, z-up\n\n[Paper](https://arxiv.org/abs/1903.11027)",
      "votes": null
    },
    {
      "id": "636051",
      "postDate": "09/28/2019 17:40:33",
      "content": "<p>So, basically both KITTI and NuScenes follow right hand rule (thumb: z, index finger: x, middle finger: y). KITTI's lidar frame is 90 degrees rotated to NuScenes' Lidar frame, can confirm this from <a href=\"https://github.com/nutonomy/nuscenes-devkit/blob/master/python-sdk/nuscenes/scripts/export_kitti.py#L92\">here</a> and <a href=\"https://github.com/nutonomy/nuscenes-devkit/blob/master/python-sdk/nuscenes/utils/kitti.py#L28\">here</a>.</p>",
      "rawMarkdown": "So, basically both KITTI and NuScenes follow right hand rule (thumb: z, index finger: x, middle finger: y). KITTI's lidar frame is 90 degrees rotated to NuScenes' Lidar frame, can confirm this from [here](https://github.com/nutonomy/nuscenes-devkit/blob/master/python-sdk/nuscenes/scripts/export_kitti.py#L92) and [here](https://github.com/nutonomy/nuscenes-devkit/blob/master/python-sdk/nuscenes/utils/kitti.py#L28).",
      "votes": null
    },
    {
      "id": "636142",
      "postDate": "09/28/2019 21:29:16",
      "content": "<p>From the annotation, it seems different. Here's an example. If we look at the bicycle (red box) in the following rendered results, it seems that x is front and y is left. <a href=\"/rishabhiitbhu\">@rishabhiitbhu</a> </p>\n\n<p><img src=\"https://www.googleapis.com/download/storage/v1/b/kaggle-user-content/o/inbox%2F1479654%2F793e09309f4bd9f612226c5882c02728%2Flidar_top.JPG?generation=1569848921118012&amp;alt=media\" alt=\"\">\n <img src=\"https://www.googleapis.com/download/storage/v1/b/kaggle-user-content/o/inbox%2F1479654%2Fdf3750e31d1e8d49047de69ef2c4d4fa%2Fxyz.JPG?generation=1569848928642793&amp;alt=media\" alt=\"\"></p>",
      "rawMarkdown": "From the annotation, it seems different. Here's an example. If we look at the bicycle (red box) in the following rendered results, it seems that x is front and y is left. @rishabhiitbhu \n\n![](https://www.googleapis.com/download/storage/v1/b/kaggle-user-content/o/inbox%2F1479654%2F793e09309f4bd9f612226c5882c02728%2Flidar_top.JPG?generation=1569848921118012&amp;alt=media)\n ![](https://www.googleapis.com/download/storage/v1/b/kaggle-user-content/o/inbox%2F1479654%2Fdf3750e31d1e8d49047de69ef2c4d4fa%2Fxyz.JPG?generation=1569848928642793&amp;alt=media)",
      "votes": null
    }
  ],
  "comments": [
    {
      "id": 632265,
      "author_name": "rishabhiitbhu",
      "author_url": "",
      "post_date": "09/23/2019 12:10:46",
      "content": "<p><img src=\"https://www.googleapis.com/download/storage/v1/b/kaggle-user-content/o/inbox%2F761152%2F8b6b2b756e577017acbb9a24c9337561%2Fpic-selected-190923-1738-51.png?generation=1569240554397179&amp;alt=media\" alt=\"\"></p>\n\n<p>Taken from the nuscenes dataset, as lyft also follows the same format we can say that for \nLIDAR_TOP: x-right, y-front, z-up</p>\n\n<p><a href=\"https://arxiv.org/abs/1903.11027\">Paper</a></p>",
      "votes": null,
      "replies": [
        {
          "id": 636051,
          "author_name": "rishabhiitbhu",
          "author_url": "",
          "post_date": "09/28/2019 17:40:33",
          "content": "<p>So, basically both KITTI and NuScenes follow right hand rule (thumb: z, index finger: x, middle finger: y). KITTI's lidar frame is 90 degrees rotated to NuScenes' Lidar frame, can confirm this from <a href=\"https://github.com/nutonomy/nuscenes-devkit/blob/master/python-sdk/nuscenes/scripts/export_kitti.py#L92\">here</a> and <a href=\"https://github.com/nutonomy/nuscenes-devkit/blob/master/python-sdk/nuscenes/utils/kitti.py#L28\">here</a>.</p>",
          "votes": null,
          "replies": []
        },
        {
          "id": 636142,
          "author_name": "yi233333",
          "author_url": "",
          "post_date": "09/28/2019 21:29:16",
          "content": "<p>From the annotation, it seems different. Here's an example. If we look at the bicycle (red box) in the following rendered results, it seems that x is front and y is left. <a href=\"/rishabhiitbhu\">@rishabhiitbhu</a> </p>\n\n<p><img src=\"https://www.googleapis.com/download/storage/v1/b/kaggle-user-content/o/inbox%2F1479654%2F793e09309f4bd9f612226c5882c02728%2Flidar_top.JPG?generation=1569848921118012&amp;alt=media\" alt=\"\">\n <img src=\"https://www.googleapis.com/download/storage/v1/b/kaggle-user-content/o/inbox%2F1479654%2Fdf3750e31d1e8d49047de69ef2c4d4fa%2Fxyz.JPG?generation=1569848928642793&amp;alt=media\" alt=\"\"></p>",
          "votes": null,
          "replies": []
        }
      ]
    }
  ],
  "raw_markdown_by_id": {
    "631949": "Hi,\n\nCould you please give us a clear axis definition of LiDAR in lyft data? For example, in kitti dataset, its axis definition of LiDAR is \"x-front, y-left, z-up\".\n\nMany thanks.",
    "632265": "![](https://www.googleapis.com/download/storage/v1/b/kaggle-user-content/o/inbox%2F761152%2F8b6b2b756e577017acbb9a24c9337561%2Fpic-selected-190923-1738-51.png?generation=1569240554397179&amp;alt=media)\n\nTaken from the nuscenes dataset, as lyft also follows the same format we can say that for \nLIDAR_TOP: x-right, y-front, z-up\n\n[Paper](https://arxiv.org/abs/1903.11027)",
    "636051": "So, basically both KITTI and NuScenes follow right hand rule (thumb: z, index finger: x, middle finger: y). KITTI's lidar frame is 90 degrees rotated to NuScenes' Lidar frame, can confirm this from [here](https://github.com/nutonomy/nuscenes-devkit/blob/master/python-sdk/nuscenes/scripts/export_kitti.py#L92) and [here](https://github.com/nutonomy/nuscenes-devkit/blob/master/python-sdk/nuscenes/utils/kitti.py#L28).",
    "636142": "From the annotation, it seems different. Here's an example. If we look at the bicycle (red box) in the following rendered results, it seems that x is front and y is left. @rishabhiitbhu \n\n![](https://www.googleapis.com/download/storage/v1/b/kaggle-user-content/o/inbox%2F1479654%2F793e09309f4bd9f612226c5882c02728%2Flidar_top.JPG?generation=1569848921118012&amp;alt=media)\n ![](https://www.googleapis.com/download/storage/v1/b/kaggle-user-content/o/inbox%2F1479654%2Fdf3750e31d1e8d49047de69ef2c4d4fa%2Fxyz.JPG?generation=1569848928642793&amp;alt=media)"
  },
  "source": "meta"
}