DeepV2D: Video to Depth with Differentiable Structure from Motion.
Zachary TeedJia DengPublished in: ICLR (2020)
Keyphrases
- structure from motion
- scene depth
- bundle adjustment
- image sequences
- point cloud
- computer vision
- camera parameters
- missing data
- motion estimates
- feature tracking
- camera motion
- depth information
- depth map
- objective function
- scene structure
- reprojection error
- feature correspondences
- focal length
- relative pose
- camera pose estimation
- image based localization
- camera positions
- least squares