Yuhong Feng

Papers

Total Citations

H-Index

About

Yuhong Feng is a leading researcher in computer vision and 3D scene understanding, with a particular focus on bridging the gap between 2D imagery and 3D spatial reasoning in complex, dynamic environments. Her most notable contribution is the development of TransCNNLoc, a pioneering end-to-end framework that leverages a hybrid Transformer-CNN architecture for pixel-level learning of 2D-to-3D pose estimation. This work, published in 2023 and already garnering 9 citations, addresses a critical challenge in robotics and augmented reality: accurately estimating camera pose in indoor scenes where objects and people are in motion. By integrating global context from Transformers with local feature extraction from CNNs, Feng’s approach achieves robust performance without relying on traditional geometric priors or depth sensors. Her research is highly impactful for applications in autonomous navigation, human-robot interaction, and mixed reality, where real-time, reliable pose estimation is essential. Feng’s work stands out for its practical elegance—offering a scalable solution that moves beyond static, controlled settings to handle the unpredictability of real-world indoor spaces.