Benzhang Wang

National University of Defense Technology

Papers

Total Citations

H-Index

About

Dr. Benzhang Wang is a computer vision researcher whose work focuses on advancing depth estimation techniques for mobile robotics and autonomous systems. His most cited contribution, "Unsupervised Stereo Depth Estimation Refined by Perceptual Loss" (2018), addresses a critical challenge in the field: achieving high-quality depth perception without expensive ground-truth labels. By integrating perceptual loss functions into unsupervised learning frameworks, Wang demonstrated that deep convolutional neural networks could rival supervised methods in accuracy while maintaining the flexibility of unsupervised approaches. This work has garnered 3 citations, reflecting its niche but foundational impact on self-supervised depth estimation pipelines. Wang’s research sits at the intersection of deep learning, 3D reconstruction, and robotic perception, where he explores how neural networks can extract spatial information from stereo imagery with minimal human annotation. His contributions are particularly valuable for real-world applications like autonomous navigation and scene understanding, where labeled data is scarce. Through his innovative use of perceptual loss, Wang has helped bridge the gap between traditional stereo algorithms and modern deep learning, offering a pathway toward more robust and scalable depth sensing systems.