Show HN: Dual YOLOv8n UAV Detection on RK3588S at 42 FPS Using NPU
Summary
Real-time UAV detection with YOLOv8n on RK3588S, running across three NPUs to reach 46 FPS at 1080p while using about 140 MB RAM per stream. The project demonstrates a fully hardware-accelerated pipeline where capture, color conversion, and inference run on fixed-function blocks (ISP, RGA, NPU) with the CPU left free. It describes a multi-process topology and provides build/deploy instructions for native and cross-compilation, plus architecture diagrams and repository structure.