Guangliang Cheng

Papers

Total Citations

211

H-Index

About

Guangliang Cheng is a leading researcher at the forefront of visual perception and robotic intelligence, whose work bridges the gap between computer vision and real-world autonomous systems. His primary research areas include visual segmentation, open-vocabulary scene understanding, and robotic manipulation. Cheng’s most impactful contribution is his comprehensive survey on transformer-based visual segmentation (2024), which has already garnered 192 citations. This seminal work systematically maps the evolution of deep learning methods for partitioning images, video frames, and point clouds—a critical capability for autonomous driving, medical imaging, and robot sensing. Beyond surveys, Cheng has made notable technical advances with OVGNet (2024, 8 citations), a unified visual-linguistic framework that enables robots to recognize and grasp novel-category objects without prior training. This work directly addresses one of robotics’ hardest challenges: generalizing to unseen objects in unstructured environments. By integrating language understanding with visual perception, Cheng is helping to create more adaptable, intelligent machines. His research continues to shape how autonomous systems perceive and interact with the world, making him a key figure in the next generation of embodied AI.