Rockchip Edge AI

Designing complete edge-AI products: camera to NPU to application, with LLMs on RK3588/RK3576.

An edge-AI product is more than a model. It needs a camera or video ingest (MPP decode for RTSP streams), pre-processing on RGA, NPU inference, CPU post-processing and an application layer (MQTT, REST, UI). RK3588 can decode many 1080p streams while running detection on all three NPU cores; RKLLM brings small LLMs on-device.

Key points

Budget memory bandwidth: video decode, RGA and NPU all share DDR.
Use zero-copy DMA buffers between MPP, RGA and RKNN where possible.
Watch thermals. Sustained NPU + VPU load needs a proper heatsink or the SoC throttles.

References

Edge AI guides

AI & NPUIntermediate

Convert YOLO to RKNN and Run It on the RK3588 NPU

Export YOLO to ONNX, quantise and convert it to .rknn with RKNN-Toolkit2 on an x86 host, then run it on the board with RKNN-Toolkit-Lite2, using all three RK3588 NPU cores.

10 min read · RK3588, RK3588S, RK3576

Spotted an error or have something to add? Suggest an improvement.

Building a Rockchip-based product?

From hardware design to Linux, Android and AI integration, get engineering support for your Rockchip project.

Request Engineering Support See what we do