TransPillars: Coarse-to-Fine Aggregation for Multi-Frame 3D Object Detection.
Zhipeng LuoGongjie ZhangChangqing ZhouTianrui LiuShijian LuLiang PanPublished in: WACV (2023)
Keyphrases
- coarse to fine
- d objects
- multi frame
- object detection
- object recognition
- multiscale
- multiresolution
- point correspondences
- multiple views
- pose estimation
- viewpoint
- multi view
- motion estimation
- image registration
- image sequences
- three dimensional
- super resolution
- shape descriptors
- dynamic programming
- object class
- computationally efficient
- motion analysis
- optical flow
- object categories
- feature correspondences
- real world objects
- computer vision
- object classes
- mean shift
- scale space
- video frames
- object segmentation
- high quality