+-------------------------------------------------------+ | User Application Code | +-------------------------------------------------------+ | Middleware (OpenCV, Live555, QT / LVGL) | +-------------------------------------------------------+ | SigmaStar MI (Media Independent Interface) Layer | +-------------------------------------------------------+ | Kernel Drivers / Hardware Abstraction (HAL) | +-------------------------------------------------------+ | SigmaStar SoC Hardware (ISP/VPU/NPU) | +-------------------------------------------------------+ Hardware Abstraction & Drivers
Controls sensor interfaces (MIPI-CSI/Sub-LVDS) and ISP parameters.
If you are looking to get started without official manufacturer support, these resources are essential:
The SDK provides extensive tools for ISP calibration, allowing engineers to fine-tune wide dynamic range (WDR), noise reduction, and low-light performance. For AI-driven tasks, the SDK includes a dedicated workflow—often involving a "Toolkit" that converts standard models (like Caffe, TensorFlow, or ONNX) into a format compatible with SigmaStar’s hardware. This enables real-time person detection, face recognition, and vehicle tracking directly on the device. Efficiency in Development
SigmaStar bypasses this through zero-copy hardware binding managed by the MI_SYS module. You tell the system to connect the Output Port of one hardware module directly to the Input Port of another hardware module in memory. Visualizing the Pipeline sigmastar sdk
MI_SYS_Init(); MI_VENC_CreateChn(0, ...); MI_RTSP_CreateSession("live", 8554);
Unlike generic Linux, you configure the kernel and drivers using a menuconfig system:
SigmaStar simplifies image generation by keeping execution logic within the project/ directory. Step 1: Selecting the Board Configuration
Always inspect the conversion logs to ensure your layers are mapping fully to the rather than falling back to software emulation on the ARM CPU. the Linux kernel
Run make config within /project directory to define target profile.
Includes the ARM Cortex CPU cores, ISP (Image Signal Processor), VPU (Video Processing Unit), and NPU (Neural Processing Unit) for AI workloads.
make sdk – Links the multimedia drivers and structures modules. Step 3: Output Images
The heart of the SDK. It manages data flow between all modules, handling channel registration, port binding, and memory management. Modular Pipeline: VIF (Video Input): Captures raw data from MIPI or DVP interfaces. VPE (Video Process Engine): Handles scaling, rotation, and image enhancement. VENC (Video Encoder): visit SigmaStarDocs .
The SigmaStar SDK provides comprehensive drivers, middleware, and tools for developing on SSD20X and SSD220 SoCs, featuring a Media Integration (MI) layer for managing data flow across modules like VDEC, ISP, and audio. Development requires setting up a cross-compilation environment for building bootloaders, the Linux kernel, and user-space applications to generate flashable images. For detailed technical documentation and guides, visit SigmaStarDocs . Environment setup - SigmaStarDocs
SigmaStar ships with an ancient arm-linux-gnueabihf-gcc 5.4.0 . If you try to use C++11 or modern pthreads, you’ll get random segfaults.
Execute the setup script to choose your specific chip and reference board layout (e.g., alkaid_ssd202_demoboard_defconfig ). make alkaid_ssd202_demoboard_defconfig Use code with caution.
Future outlook
Extract the toolchain provided in your SDK distribution package to /opt/sigmastar/ .