Junpei Huang

University of Science and Technology of China

Papers

Total Citations

H-Index

About

Junpei Huang is a rising researcher at the forefront of efficient computing for embodied intelligence, with a primary focus on robotics and computer vision. His work addresses the critical challenge of enabling real-time, low-power perception and control in resource-constrained robotic and autonomous systems. Huang’s major contributions lie in bridging the gap between complex algorithms and practical hardware. In his highly cited work, "Dadu-SV," he pioneered a method to accelerate stereo vision processing—a core task for depth perception—on Neural Processing Units (NPUs), demonstrating how to efficiently implement both classic Semi-Global Matching (SGM) and deep CNNs. This work has garnered 4 citations for its practical approach to a pervasive bottleneck. Complementing this, his tutorial "Toward Efficient Computing for Robotics" provides a comprehensive circuit- and system-level guide to designing accelerators for kinematics, motion planning, and perception. With a total of 6 citations on his most prominent papers, Huang is establishing himself as a key voice in the hardware-software co-design community, making advanced robotic capabilities more accessible and energy-efficient for next-generation autonomous machines.