fudan-zvg / PolarFormer

[AAAI 2023] PolarFormer: Multi-camera 3D Object Detection with Polar Transformers
MIT License
160 stars 15 forks source link

same question about Cross-plane encoder #13

Open yangH8 opened 1 year ago

yangH8 commented 1 year ago

In the self._forward_single_camera, you align the image column and polar ray with cross-attention, but there is no use of cam2lidar and cam_intrinsic, this does not correspond to the description in the Cross-plane encoder section of the paper. Please add to this section.