Papers
1
Total Citations
1
H-Index
1
About
Yuzhe Fu is a rising star in the field of energy-efficient deep learning hardware, with a primary focus on accelerating point-cloud neural networks (PNNs) for 3D perception systems. His most notable contribution is the development of **Nebula**, a 28nm 3D PNN accelerator that achieves a remarkable **109.8 TOPS/W**—one of the highest reported energy efficiencies for point-cloud processing. This work introduces three key innovations: adaptive partition for dynamic workload balancing, multi-skipping to eliminate redundant computations, and block-wise aggregation for efficient memory access. These techniques directly address the computational bottlenecks of point-based PNNs, which are critical for real-time applications in autonomous driving, robotics, and virtual reality. Though early in his career, Fu’s work has already garnered attention for its practical impact on deploying 3D neural networks in resource-constrained edge devices. His research bridges the gap between algorithmic advances in point-cloud analysis and the hardware efficiency required for real-world deployment, positioning him as a promising contributor to the next generation of intelligent sensing systems.
Research Focus
Key Achievements
Top Papers
- 1