Accelerating Robotics Perception on AMD Ryzen AI APU
COMP5709 Capstone Project — University of Sydney
Project Overview
Robotics systems require fast and efficient perception to operate in real-world environments.
This project explores how AMD Ryzen AI APU can accelerate Vision-Language Models (VLMs) using:
- CPU
- Integrated GPU (iGPU)
- Neural Processing Unit (NPU)
We focus on two models: - VGGT - SmolVLM
Experimental Platform
The experiments are conducted on a compact edge device:
Beelink SER Mini PC
- AMD Ryzen 7 8845HS
- 32 GB RAM
- 1 TB SSD