Show HN: Dual YOLOv8n UAV Detection on RK3588S at 42 FPS Using NPU
42 FPS on edge. YOLOv8n dual inference on RK3588S NPU shows the real constraint isn't the model—it's getting it to run where it matters.
Why it matters
Edge AI inference optimization on specialized hardware (NPU) demonstrates practical deployment challenges beyond model capability—relevant to founders building real-world vision systems and infrastructure teams evaluating embedded inference strategies.
The key facts
12 to knowDual YOLOv8n inference at 42 FPS
RK3588S NPU hardware
Multi-threaded implementation
Open-source GitHub project
Edge deployment (not cloud-based)
UAV detection use case
Dual YOLOv8n models running simultaneously at 42 FPS
RK3588S SoC with NPU acceleration
Open-source implementation published on GitHub
UAV detection use case (drone/robotics application)
Edge inference without cloud dependency
Published June 2026 (Hacker News community project)
Go to the source
Hacker Newsgithub.com
Publisher excerpt: Article URL: Comments URL: Points: 19 # Comments: 2