ROAM: Robust and Object-aware Motion Generation using Neural Pose Descriptors.
Wanyue ZhangRishabh DabralThomas LeimkühlerVladislav GolyanikMarc HabermannChristian TheobaltPublished in: CoRR (2023)
Keyphrases
- object pose
- partial occlusion
- rigid objects
- position and orientation
- d objects
- object representation
- pose estimation
- articulated objects
- cluttered background
- robust tracking
- initial position
- illumination variations
- moving objects
- object tracking
- keypoints
- feature descriptors
- tracking framework
- deformable objects
- object motion
- feature vectors
- target object
- perspective projection
- long video sequences
- object shape
- multibody
- bounding box
- image sequences
- space time
- pose space
- object segmentation
- neural network
- motion segmentation
- shape descriptors
- lighting conditions
- network architecture
- motion estimation
- optical flow
- video objects
- dynamic scenes
- motion model
- feature points
- appearance variations
- object appearance
- multi view
- visual motion