How it works:
1. Estimate 3D surface orientations.
2. Detect cars and people and estimate viewpoint.
3. Recover occlusion boundaries.
4. Repeat, using all available information to refine estimates.
See our papers for details:
www.cs.uiuc.edu/homes/dhoiem/publications/index.html
Derek Hoiem, Alexei A. Efros, Martial Hebert, Closing the Loop on Scene Interpretation, CVPR 2008.
Link to this comment:
All Comments (0)