Website arXiv Video

Automatically generate realistic training data from time-lapse videos for reconstructing dynamics objects (people, vehicle) under occlusion.

Website arXiv Code Video

3D scene reconstruction and precise localization of over 100 real-world traffic cameras.

Website arXiv Code Video

Extension of the spatial rectifier to the multi-directional case, applying to depth/surface normal prediction from egocentric view, accompanied by a novel egocentric RGB-D dataset.

arXiv Code

Depth estimation by multiview triangulation, followed by an iterative depth refinement module that preserves estimates with high triangulation confidence.

Website arXiv Code Video

Robust surface normal estimation by spatially rectifying image to the densely distributed orientations.

arXiv Code

Depth estimation by completing a very sparse VI-SLAM point cloud using planar constraint from robust surface normal prediction.