Self-Supervised Structure-from-Motion through Tightly-Coupled Depth and Egomotion Networks.
Brandon WagstaffValentin PeretroukhinJonathan KellyPublished in: CoRR (2021)
Keyphrases
- structure from motion
- ego motion
- tightly coupled
- camera motion
- image sequences
- scene structure
- motion stereo
- bundle adjustment
- fine grained
- optical flow
- loosely coupled
- d scene
- camera parameters
- point cloud
- field of view
- motion parameters
- general purpose
- feature tracking
- epipolar geometry
- stereo camera
- moving objects
- depth map
- moving camera
- video sequences
- depth information
- computer vision
- missing data
- focal length
- robust estimation
- motion field
- dynamic scenes
- object tracking
- feature points
- flow field
- vector field
- camera pose
- motion model
- motion estimation
- least squares
- feature correspondences