Show HN: Dual YOLOv8n UAV Detection on RK3588S at 42 FPS Using NPU

Article excerpt

A developer shared an open-source project demonstrating dual YOLOv8n object detection running on a Khadas board with RK3588S processor, achieving 42 frames per second using the chip's neural processing unit. The work showcases how to run two simultaneous AI vision models for detecting unmanned aerial vehicles on edge hardware, exactly the kind of resource-constrained deployment that powers embedded computer vision applications. The GitHub repository includes code for multithreading the inference across the NPU, suggesting a practical solution for real-time drone detection on modest hardware.