Convert map frame 3D Box label into velo frame 3D Box

[lejk8104]

Thank you for your contributions

I am training 3D object Detection model in KITTI-360 training dataset (only use 3D bounding box data)

I am parsing data 3D bboxes data, but I have some problems,

velodyne pointcloud data is lidar_frame(sensor coordinate) but, 3D bounding box is map_frame(global coordinate)

How to convert it? How parsing and extract "poses.txt "?

Have you some reference?

thanks

[yiyiliao]

Hi, in our leaderboard, the baselines are trained on the accumulated point cloud (in global coordinate) instead of the per-frame LiDAR points.

If you are interested in the per-frame setting, you can convert the LiDAR points to the global coordinate following this implementation.