Show HN: Dual YOLOv8n UAV Detection on RK3588S at 42 FPS Using NPU (github.com) AI

The GitHub project describes a real-time UAV detection pipeline for Rockchip RK3588S boards that uses hardware-accelerated camera/ISP and RGA for pre-processing plus YOLOv8n inference on the device’s three NPU cores, achieving up to the camera’s ~46 FPS ceiling (~140–150 MB RAM per 1080p stream). It also tracks detections with ByteTrack and generates natural-language summaries using an on-device Qwen2.5-0.5B LLM when UAVs leave the scene, streaming annotated output to HDMI or RTSP.