James Tu
Papers
1
Total Citations
8
H-Index
1
About
James Tu is a leading researcher in robotics perception, with a focus on efficient, real-time 3D object detection from LiDAR data. His most notable contribution is the development of **StrObe**, a novel framework for streaming object detection directly from raw LiDAR packets, bypassing the traditional and computationally expensive step of accumulating full 360° point clouds. This work, published in 2020, addresses a critical bottleneck in autonomous systems: the latency introduced by waiting for a complete sensor sweep. By processing data sector-by-sector as it arrives, StrObe enables faster reaction times for safety-critical applications like autonomous driving. The paper has garnered 8 citations, establishing Tu as an early innovator in streaming perception—a paradigm that prioritizes low-latency, asynchronous processing over batch accumulation. His research bridges the gap between sensor hardware constraints and algorithmic design, demonstrating how to exploit the natural packet structure of rolling shutter LiDARs for more responsive robotic systems. Tu’s work is essential reading for students and engineers seeking to push the boundaries of real-time 3D vision in resource-constrained, time-sensitive environments.
Research Focus
Key Achievements
Top Papers
- 1StrObe: Streaming Object Detection from LiDAR Packets8 citations · 2020