Open Kishaan opened 9 months ago
Hi, very valid question, we need to clarify this in the documentation :)
yaw = 0 corresponds to a vehicle that is oriented the same way as the ego-vehicle. This means that it follows the lidar coordinate system of the ego-vehicle meaning x-right, y-forward, z-up. Consequently the assignment of dimension to axis is: x -> width y -> length z -> height
Does the z axis of the 3D bbx provide the depth of the object with respect to the sensor as it follows the lidar coordinate?
I think the notion of depth is quite ambiguous in the context of a 360 degree surround lidar. However, if you project the point into the camera coordinate system, then the z coordinate would indeed be considered depth.
If working in the context of lidar, you probably want euclidean distance instead (e.g. np.linalg.norm(point_cloud[:,3])
)
Hi,
Thank you for releasing the dataset. I would like to know whether the length, width, and height of the 3D Bboxes available in the annotations correspond to x, y, and z axes of the lidar coordinate system, respectively. If not, please provide some clarity with respect to which axes the length, width, and height are annotated.
Thank you!