Yiliu Feng

National University of Defense Technology

Papers

Total Citations

H-Index

About

Yiliu Feng’s research lies at the intersection of computer vision and robotics, with a primary focus on advancing depth estimation techniques for autonomous systems. Her most-cited work, “Unsupervised Stereo Depth Estimation Refined by Perceptual Loss” (2018), tackles a fundamental challenge: enabling mobile robots to perceive three-dimensional environments without the costly, labor-intensive ground-truth data required by supervised learning methods. By integrating perceptual loss into an unsupervised deep convolutional neural network framework, Feng demonstrated that machines could learn depth from binocular images with significantly improved accuracy and robustness—closing the gap between unsupervised and supervised approaches. This contribution has garnered 3 citations and serves as a stepping stone for more efficient, scalable depth perception in real-world robotics applications. Feng’s work is particularly notable for its practical orientation, addressing the data scarcity problem that often limits deployment in dynamic, unstructured environments. Her research continues to influence the development of cost-effective, learning-based solutions for autonomous navigation, object detection, and scene understanding, making her a promising voice in the push toward truly self-sufficient robotic vision systems.