3D Hierarchical Refinement and Augmentation for Unsupervised Learning of Depth and Pose from Monocular Video.
Guangming WangJiquan ZhongShijie ZhaoWenhua WuZhe LiuHesheng WangPublished in: CoRR (2021)
Keyphrases
- unsupervised learning
- pose estimation
- pose tracking
- depth cues
- depth video
- monocular images
- supervised learning
- multimedia
- video data
- depth perception
- depth images
- video streams
- video sequences
- visual cues
- pose recovery
- video frames
- expectation maximization
- human pose
- dimensionality reduction
- image sequences
- real time
- video content
- semi supervised
- computer vision
- person detection
- deep learning
- ego motion
- motion capture
- key frames
- articulated human motion
- stereo camera
- position and orientation
- depth information
- depth map
- d objects
- depth estimation
- dynamic scenes
- space time
- human body
- free viewpoint
- text classification
- optical flow
- deep architectures
- feature selection
- machine learning