Yunsong Zhou
Papers
3
Total Citations
43
H-Index
2
About
Yunsong Zhou is an emerging researcher specializing in computer vision and autonomous perception, with a particular focus on monocular 3D object detection (Mono3D) for mobile platforms. His work addresses one of the field's most demanding challenges: enabling accurate, real-time 3D scene understanding using only a single camera on resource-constrained devices such as vehicles, drones, and robots. Zhou's most notable contribution, **MonoATT** (2023), introduces an Adaptive Token Transformer that intelligently allocates computational resources by moving beyond rigid grid-based vision tokens, allowing efficient yet precise online detection — a significant step forward for deployment on edge hardware. This work has attracted 37 citations, signaling strong community interest. Complementing this, his **MoGDE** framework tackles the near-far disparity problem inherent in monocular vision by leveraging ground depth estimation to improve detection accuracy for distant objects under changing camera poses — a critical challenge in real-world mobile applications. Collectively, Zhou's research pushes the boundary of what is achievable with lightweight, single-camera perception systems, making meaningful contributions to safer and more capable autonomous platforms. His growing citation record reflects the practical relevance and technical rigor of his early-career output.
Research Focus
Key Achievements
Top Papers
- 1
- 2
- 3