BEVFormer
BEVFormer copied to clipboard
Issue of Lidar2ego and ego2global translation?
Hi @zhiqi-li
During calculate can_bus[:3] and can_bus[-2:], the 'ego2global_rotation' and 'ego2global_translation' are used. However, as the BEV feature (and GT labels) are on Lidar coordinate, should 'lidar2ego' also used in can_bus calculation?
I also have the same question.
@zhiqi-li Thanks for your sharing. Could you explain it to me, because the label is in the LiDAR coordinate system, but the features extracted from the image are in the Camera coordinate system, where the coordinate conversion is performed?Thank you very much!