Computer Vision based Co-Robot Wheelchair
I am now working on this project, in which we build an intelligent assistive robotic wheelchair and develop convenient computer vision based method for hands-free mobility. We build this robotic wheelchair from a powered wheelchair. It is a lot of fun!
Fast Multi-view Face Detection
I have been working on face detection during my two internships with Adobe Research. We develop fast multi-view face detectors for real-world scenario. We use deep learning to learn discriminative features and design a cascade of Convolutional Neural Networks (CNN) for acceleration. Our face detector achieves top-level performance on the FDDB benchmark and runs fast even with single CPU core.
Real-world Face Recognition
This is my PhD research topic. We develop the Probabilistic Elastic Part (PEP) model for the real-world face recognition and face detector adaptation. Our methods achieved top-level performance on many face recognition benchmarks including Labeled Faces in-the-Wild (LFW), YouTube video faces dataset and PaSC Face Recognition Challenge.