Regulating Intermediate 3D Features for Vision-Centric Autonomous Driving
This addresses a bottleneck in vision-centric autonomous driving perception by improving feature extraction for multiple tasks, though it appears incremental as it builds on existing LSS frameworks.
The paper tackles the problem of generating unsuitable dense 3D features in multi-camera autonomous driving perception by proposing Vampire, which regulates intermediate features using volume rendering, resulting in competitive performance with SOTA methods on tasks like 3D occupancy prediction and 3D object detection on Occ3D and nuScenes datasets.
Multi-camera perception tasks have gained significant attention in the field of autonomous driving. However, existing frameworks based on Lift-Splat-Shoot (LSS) in the multi-camera setting cannot produce suitable dense 3D features due to the projection nature and uncontrollable densification process. To resolve this problem, we propose to regulate intermediate dense 3D features with the help of volume rendering. Specifically, we employ volume rendering to process the dense 3D features to obtain corresponding 2D features (e.g., depth maps, semantic maps), which are supervised by associated labels in the training. This manner regulates the generation of dense 3D features on the feature level, providing appropriate dense and unified features for multiple perception tasks. Therefore, our approach is termed Vampire, stands for "Volume rendering As Multi-camera Perception Intermediate feature REgulator". Experimental results on the Occ3D and nuScenes datasets demonstrate that Vampire facilitates fine-grained and appropriate extraction of dense 3D features, and is competitive with existing SOTA methods across diverse downstream perception tasks like 3D occupancy prediction, LiDAR segmentation and 3D objection detection, while utilizing moderate GPU resources. We provide a video demonstration in the supplementary materials and Codes are available at github.com/cskkxjk/Vampire.