FD3D: Exploiting Foreground Depth Map for Feature-Supervised Monocular 3D Object Detection.
Zizhang WuYuanzhu GanYunzhe WuRuihao WangXiaoquan WangJian PuPublished in: AAAI (2024)
Keyphrases
- d objects
- depth map
- multi view
- image sequences
- pose estimation
- three dimensional
- depth cues
- depth information
- d scene
- stereo camera
- multiple views
- viewpoint
- stereo matching
- object recognition
- range data
- view synthesis
- depth estimation
- stereo pair
- d mesh
- depth images
- object boundaries
- object detection
- semi supervised
- photometric stereo
- high quality
- dynamic scenes
- range images
- feature vectors
- optical flow
- surface patches
- computer vision
- camera calibration
- input image
- image features
- shape descriptors
- stereo vision
- object tracking
- video sequences
- vision system