OpenNVR 0.1.10: running an open-source NVR's AI on the Qualcomm Hexagon NPU

说明与用途

OpenNVR 0.1.10 采用 Qualcomm Hexagon NPU，将 YOLO 推理时间降至 8.6 ms，并支持 16 台摄像头 2 fps 无帧丢失。

问题

Deploying AI inference for video surveillance on a low‑power, air‑gapped device while maintaining performance.

结果

Using the Qualcomm Hexagon NPU, OpenNVR achieves a 19× speed‑up for YOLO inference and supports 16 cameras at 2 fps without frame drops, compared to 4 cameras on CPU.

工作流

• Clone OpenNVR repo
• Run start.sh with OPENNVR_ACCEL=auto to auto‑detect Qualcomm NPU
• Verify provider used via metrics
• Benchmark with detect_pipeline.bench for latency and provider confirmation

证据

Tier-0 detector (YOLO, one frame): CPU, 8 threads 164 ms; Hexagon NPU, fp16 8.6 ms

来源

In the full stack (streams, decode, tracking, the API, all on the same box): CPU 4 cameras at 2 fps, no drops; NPU 16, 0 dropped

来源

Everything goes through ONNX Runtime's QNN execution provider, shipped as a plugin wheel (onnxruntime-qnn). There is no Qualcomm SDK to install on the host.

来源