Back to Basics: Unsupervised Learning of Optical Flow via Brightness Constancy and Motion Smoothness.
Jason J. YuAdam W. HarleyKonstantinos G. DerpanisPublished in: ECCV Workshops (3) (2016)
Keyphrases
- optical flow
- brightness constancy
- unsupervised learning
- motion boundaries
- motion model
- brightness constancy assumption
- optical flow computation
- optical flow estimation
- motion field
- smoothness constraint
- spatial smoothness
- motion estimation
- image sequences
- flow estimation
- motion parameters
- flow field
- optical flow algorithm
- image motion
- camera motion
- moving objects
- motion segmentation
- velocity field
- motion analysis
- computer vision
- d scene
- consecutive frames
- feature tracking
- multi frame
- space time
- ego motion
- moving camera
- semi supervised
- multiscale
- vector field
- model selection
- rigid motion
- image brightness
- expectation maximization
- machine learning
- object recognition