Ported OpenHarmony 5.0.3 to the Huawei Ascend 310B developer kit, then built Ascend X-Ray Expert — a HarmonyOS app that runs YOLOv5s on the NPU to analyze chest radiographs. From kernel patch adaptation to ArkUI, this project spans the complete embedded system stack.
Qizhi01 covers the full chain from operating system to application: porting OpenHarmony 5.0.3 to the Huawei Ascend 310B development kit, then building an X-ray chest radiograph analysis application on top of that system. The first half is kernel, driver and build-chain adaptation; the second half is an ArkTS interface with YOLOv5s inference. To put the on-board NPU to actual use, the application implements both a CPU and an NPU inference path.
The Ascend 310B ships with a Linux SDK only, while the on-board NPU is worth using. The project verifies two things: whether OpenHarmony can run on this hardware, and whether NPU inference can be wired into a usable application on top of it.
GPU support is implemented at the HAL layer on the open-source Panfrost driver; HDMI and dual-display output are working; USB/MIPI camera and ALSA audio are connected. bash and busybox were also ported to supply the tools the HDK installer depends on.
The system side is OpenHarmony 5.0.3 compiled together with around 20 kernel modules. The application side uses ArkTS and ArkUI, with inference running through MindSpore Lite (CPU) and CANN ACL (NPU) on a YOLOv5s model.
The 43 patches are merged into the OpenHarmony mainline; the application runs on the board and completes five-class inference, with both the CPU and NPU paths verified. The X-ray classification is a technical demonstration and is not intended for clinical use.
Our lab has a Qizhi01 developer kit equipped with a Huawei Ascend 310B NPU. Out of the box it only came with a Linux SDK — I wanted to run OpenHarmony on it and put that NPU to real use. The project naturally split into two layers: the bottom layer is a system-level port of OpenHarmony 5.0.3, and the top layer is Ascend X-Ray Expert — a HarmonyOS app built with ArkTS that uses YOLOv5s inference to detect conditions like pneumonia and tuberculosis in chest X-rays. Two inference paths were implemented: MindSpore Lite on CPU and CANN ACL on NPU, both verified and functional.
The Ascend 310B chipset differs from any officially supported OpenHarmony board, requiring adaptation from the ground up
Leveraged Mali GPU's open-source Panfrost driver, implemented HAL against OpenHarmony's Display Composer VDI interface
HDMI output with dual-screen support, involving framebuffer management and VSync synchronization
V4L2 drivers for USB and MIPI CSI cameras, ALSA audio and HDMI audio output
Set up CANN inference, validated YOLOv5 models, and successfully ran MindIE Qwen3-0.6B LLM on-device
build.sh provides one-click compilation: pull OpenHarmony source → apply 43 patches → full buildformatSD310B.sh SD flashing tool for direct Linux flashing or img generation for Windows BalenaEtcherHarmonyOS 5.0.3 (API15) · ArkTS + ArkUI · Dual inference engines
YOLOv5s object detection on chest X-rays, classifying into five categories. Medical imaging is a natural fit for edge NPU inference — hospitals cannot send patient data to the cloud due to privacy and latency constraints.
Both CPU and NPU inference paths were implemented as part of the full-stack exercise:
Uses HarmonyOS's built-in MindSpore Lite Kit to load .ms model files, 2-thread CPU inference.
Preprocessing (RGBA → float normalization) and postprocessing (NMS, IoU) all implemented in ArkTS.
C++ side: dlopen loads libacl_runtime.so, then aclInit → aclrtSetDevice → aclrtCreateContext → aclrtCreateStream, with aclmdlLoadFromFile loading the .om offline model.
Exposes initYoloEngine and runYoloInference to ArkTS via NAPI.
PhotoViewPicker → resize to 640×640 with letterbox padding
RGBA → float32 → normalize → feed to inference engine
NPU via CANN ACL, CPU via MindSpore Lite
NMS + IoU → filter low-confidence overlapping detections
Canvas renders detection boxes, three-column layout — image, stats, history
Complete four-layer stack from hardware to application
Ascend 310B
ARM CPU + Mali GPU
NPU AI Accelerator
Linux 5.10
panfrost + svm
20+ .ko modules
OpenHarmony 5.0.3
Display Composer VDI
Gralloc / Audio / V4L2
ArkUI + ArkTS
Ascend X-Ray Expert
CANN NPU + MindSpore Lite
From kernel patches and HAL adaptation to ArkUI application development — the complete embedded system chain.
MindSpore Lite (CPU) and CANN ACL (NPU) both verified, with the NPU offering significantly lower inference latency.
X-ray analysis demands privacy and low latency — edge NPU inference addresses both requirements naturally.
Successfully ran Qwen3-0.6B via MindIE on the Ascend 310B, validating small LLM viability on embedded NPUs.
build.sh automates source pull, patch application, and full compilation for streamlined reproduction.
C++ CANN inference engine seamlessly integrated into the ArkTS application layer via NAPI.