My research is on computational sensing and machine learning. In particular, I am interested in combining optics and computation to measure depth and other scene information, and using machine learning techniques to assist the design process. My research thread is shown below.

Low Power Depth Sensing

We are developing efficient depth recovery algorithms and optical systems that could be used in small sensing platforms.

Focal Track
Focal Track
We built a monocular, non-active lighting, and computationally efficient depth camera, that is based on defocus created by a deformable lens. The algorithm is a derivation from focal flow. Comparing to focal flow, this depth camera works in an extended range, and runs at 100 frames per second. (ICCV 2017).
Focal Flow
We presented the focal flow algorithm, that gives local depth and 3D velocity by simply solving a linear system which only contains image derivatives. The algorithm only requires a passive, unactuated and monocular camera. We made a prototype based on the algorithm. (ECCV 2016, IJCV 2017).