Han-Pang Chiu

Technical Director, Center for Vision Technologies, Vision and Robotics Laboratory

Han-Pang Chiu, Ph.D., Technical Director of the Scene Understanding and Navigation (SUN) Group at SRI International, leads government and commercial projects to develop innovative solutions for real-world applications to multi-sensor navigation, multi-modal target detection/classification, robotic autonomy, and mobile augmented reality.

Han-Pang has been chief scientist and technical lead in many DARPA, ONR, and US Army research programs. He has been a pioneer in factor graph formulation for multi-sensor navigation. The paper “Stable Vision-Aided Navigation for Large-Area Augmented Reality” co-authored and presented by him received the best paper award in the IEEE Virtual Reality 2011 conference. His work also supports a few spin-off companies from SRI. Recently, he has also been leading SRI’s efforts in semantic navigation, which develops new deep learning techniques to derive and utilize high-level semantic scene information for robotic autonomy.

Prior to joining SRI, Han-Pang was a postdoctoral researcher in Computer Science and Artificial Intelligence Laboratory (CSAIL) at Massachusetts Institute of Technology (MIT), where he worked on a DARPA-funded project to develop computer vision techniques for robot grasping. He received his Ph.D. in Computer Science from MIT in 2009, and holds an MBA degree in management information systems from National Taiwan University.

Recent publications

more +

Incremental Learning with Differentiable Architecture and Forgetting Search (7/14/2022) - In this paper, we show that leveraging NAS for incremental learning results in strong performance gains for classification tasks.

Striking the Right Balance: Recall Loss for Semantic Segmentation (5/18/2022) - Class imbalance is a fundamental problem in computer vision applications such as semantic segmentation.

Graph Mapper: Efficient Visual Navigation by Scene Graph Generation (5/18/2022) - We propose a method to train an autonomous agent to learn to accumulate a 3D scene graph representation of its…