ECE Distinguished Speaker Seminar
Recent developments in deep representation-based methods for many computer vision problems have knocked down many research themes pursued over the last four decades. In this talk, I will discuss methods based on deep representations, adversarial learning and domain adaptation for designing robust computer vision systems with applications in unconstrained face and action verification and recognition, expression recognition, subject clustering and attribute extraction. The face and action recognition system being built at UMD is based on fusing multiple deep convolutional neural networks (DCNN) trained using publicly available still and video face data sets. Concepts such as multi-task learning, deep dictionaries, alignment-free methods and optimal sampling of training data have contributed to the design of a robust face analytics system. I will then discuss some new results on generative adversarial learning and domain adaptation for improving the robustness of the face recognition system. I will conclude the talk by discussing issues such as incorporating geometry and invariances in deep learning methods.