Dongju Li

About

Dongju Li is a researcher at the forefront of efficient deep learning hardware-software co-design, with a primary focus on deploying advanced computer vision models on resource-constrained platforms. His most notable contribution is a pioneering end-to-end implementation of the YOLOv8 object detection framework on a RISC-V architecture, achieving a power-efficient, runtime-configurable deep neural network accelerator. This work, published in 2023, directly addresses the critical challenge of bringing state-of-the-art vision capabilities—like real-time object detection—to edge devices, where power and computational resources are limited. By demonstrating that a popular and demanding framework like YOLOv8 can be effectively ported to an open-source instruction set architecture, Li’s research paves the way for more accessible, customizable, and energy-efficient AI systems in applications ranging from autonomous drones to smart sensors. While his work is still emerging, with his flagship paper already garnering 4 citations, its practical significance and alignment with the growing demand for edge AI signal a strong trajectory. Li’s contributions are essential for students and engineers seeking to bridge the gap between high-performance deep learning and the realities of embedded systems.

Research Focus

Key Achievements

Top Papers

1. 1
   A Power-efficient end-to-end Implementation of YOLOv8 Based on RISC-V

   4 citations · 2023

Key Collaborators

Contact & Links