Hello,
I am sharing a small collection of RK3588-focused projects developed and tested on
Rock 5B / Rock 5B+ boards, with a focus on RKNN-based NPU acceleration and practical,
reproducible edge-AI pipelines.
Project notes and documentation:
Included projects
Stable Diffusion 1.5 (RKNN + Euler)
-
Fully NPU-based inference (no GPU / CUDA)
-
Text-to-image, img2img, masked inpainting
-
Real-ESRGAN x4 upscaling (RKNN only)
-
rkimg command-line interface and Flask-based WebUI
-
Models and weights hosted separately on Hugging Face
Source code:
Models and assets:
RKNN Face Recognition Pipeline
-
High-performance face detection and recognition for Rockchip RK3588
-
Designed for bare-metal Flask deployment on RK3588-class boards
-
Suitable for edge and local AI workloads
Source code:
Canon EOS Studio Remote
-
Browser-based Canon EOS studio controller for Raspberry Pi and Rockchip boards
-
Remote camera control and live view for SBC-based workflows
Source code:
Additional resources
All repositories:
Platform
-
Boards: Rock 5B / Rock 5B+
-
SoC: RK3588 / RK3588S
-
Runtime: RKNN Toolkit / RKNN Toolkit Lite 2
Development note
The codebase was developed with extensive assistance from AI tooling, with manual integration,
testing, and validation performed on real RK3588 hardware. All shared projects represent
working, tested configurations.
Feedback and testing results on other RK3588-based boards are welcome.