multi-tasks in nuscenes dataset

[Liaoqing-up]

Hello, tianweiy. I would like to ask a question that has been bothering me for a long time. The common practice in the nusecens dataset is to divide the ten categories into six tasks, how does it gain over the direct classification? Have you ever done the experiment and can you share some conclusions? https://github.com/tianweiy/CenterPoint/blob/d3a248fa56db2601860d576d5934d00fee9916eb/configs/nusc/voxelnet/nusc_centerpoint_voxelnet_0075voxel_fix_bn_z_flip.py#L7-L14