无人驾驶CVPR 20252025
BEVFormer++: Improving BEV Perception with Temporal
Yuexin Ma, Tai Wang, Xuyang Bai, Huitong Yang, Yuenan Hou, Yaming Wang, Yu Qiao, Ruigang Yang, Dinesh Manocha, Xinge Zhu et al.Shanghai AI Lab
摘要
BEVFormer++ improves bird's-eye-view perception with enhanced temporal modeling and multi-scale feature fusion for more robust autonomous driving.
BEV perceptiontemporal modelingmulti-scale featuresrobust drivingfeature fusion
技术细节
模型骨架
BEVFormer++ Enhanced Transformer
编码器
Spatial/Temporal Cross-Attention Encoder + ResNet-50
解码器
Multi-Task BEV Decoder (8 layers)