Closed hutao568 closed 1 year ago
Always follow this format: https://github.com/tusen-ai/SST/blob/main/mmdet3d/core/bbox/structures/lidar_box3d.py#L15C16-L15C16 when you call the box-related torch ops.
Special reminder: the boxes in waymo_infos_xxx.pkl is not in that format, but they will be converted to the LiDARInstance3DBbox after data pipeline.
Please reopen this issue if you need further discussion.
I'm not sure whether the size dimension of the box should be l, w, h or w, l, h; when you initialize, you use x_size, y_size, z_size; but when you crop the point cloud, you use w, l , h, ry; I want to know which format should be given when the box is initialized; (here w refers to width, h refers to height)
the code for init https://github.com/tusen-ai/SST/blob/main/mmdet3d/core/bbox/structures/base_box3d.py#L19 the code for crop point cloud https://github.com/tusen-ai/SST/blob/main/mmdet3d/ops/roiaware_pool3d/points_in_boxes.py#L6