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

Category: ai-ml

Tags: uav-detection, yolov8, rk3588, npu, real-time, edge-computing, c++

Score: 7.0/10 (Innovation: 7, Technical: 8, Documentation: 7, Utility: 6)

This project implements real-time UAV detection using YOLOv8n on the RK3588S NPU, achieving 46 FPS by parallelizing inference across all 3 NPU cores and offloading all heavy processing to fixed-function silicon. It is notable for its extremely low memory footprint (~140 MB per stream) and composable pipeline that includes tracking, temporal feature extraction, and an on-device LLM for natural-language event summaries.

Target audience: embedded engineers, computer-vision engineers, edge-AI developers

Repository: https://github.com/alebal123bal/khadas_yolov8n_multithread · C++ · Apache-2.0

View on Hacker News