TransPillars: Coarse-to-Fine Aggregation for Multi-Frame 3D Object Detection.
Zhipeng LuoGongjie ZhangChangqing ZhouTianrui LiuShijian LuLiang PanPublished in: CoRR (2022)
Keyphrases
- d objects
- coarse to fine
- multi frame
- object detection
- object recognition
- multiscale
- viewpoint
- multiresolution
- multi view
- multiple views
- point correspondences
- three dimensional
- pose estimation
- motion estimation
- shape descriptors
- image registration
- super resolution
- feature correspondences
- dynamic programming
- optical flow
- image sequences
- motion analysis
- optic flow
- object class
- image processing
- machine learning
- real world objects
- object classes
- computationally efficient
- computer vision