Yusong Tan

National Defense University, National University of Defense Technology

Papers

Total Citations

H-Index

About

Yusong Tan is a researcher advancing the frontiers of computer vision and robotics, with a primary focus on stereo matching and domain adaptation. His work addresses critical challenges in enabling machines to perceive and navigate their environments with greater accuracy and efficiency. Tan’s most cited paper, “Multi-Scale Cost Volumes Cascade Network for Stereo Matching” (2021, 9 citations), tackles the fundamental trade-off between speed and accuracy in depth estimation for robot navigation. By introducing a novel cascade network that leverages multi-scale cost volumes, his method bridges the gap between traditional low-accuracy approaches and computationally expensive deep learning models, offering a more practical solution for real-time robotic systems. In his subsequent work, “Multi-Level Consistency Learning for Source-Free Model Adaptation” (2022, 3 citations), Tan addresses the pressing issue of model robustness in dynamic environments. He proposes a self-training framework that mitigates overfitting to noisy labels, enabling models to adapt to new data distributions without accessing original training data—a crucial capability for deploying robots in unpredictable real-world settings. Through these contributions, Tan is shaping more reliable and efficient perception systems for autonomous robotics.