BEVUDA: Multi-geometric Space Alignments for Domain Adaptive BEV 3D Object Detection.
Jiaming LiuRongyu ZhangXiaoqi LiXiaowei ChiZehui ChenMing LuYandong GuoShanghang ZhangPublished in: ICRA (2024)
Keyphrases
- d objects
- multi view
- object recognition
- geometric models
- shape descriptors
- recognition and pose estimation
- viewpoint
- topological information
- pose estimation
- multiple views
- three dimensional
- cad models
- spin images
- range data
- partial matching
- object classes
- d mesh
- geometric structure
- line drawings
- detection method
- object parts
- minimum bounding
- point distribution
- physically plausible
- image contours
- cad model
- pairwise
- integral imaging
- multi views
- object features
- surface patches
- object model