MCU

Nuvoton introduces NuML Toolkit for AI model deployment on MCU chips

Nuvoton Technology has launched the NuML Toolkit, a tool designed to address the challenge of deploying trained AI models on resource-constrained microcontrollers in embedded systems. Optimized for the NuMicro® M55M1 microcontroller platform, the toolkit integrates model conversion, project generation, deployment, and debugging processes.
The NuML Toolkit provides the following features:
Model Conversion and Deployment: Converts Full-INT8 quantized models into .cpp files and integrates them into Arm® Keil® project templates.

Toolchain Support: Compatible with Keil (μVision5 / Arm Compiler 6) and GNU Compiler Collection (make / gcc).

Integration: Automatically opens Keil projects and performs flashing post-deployment for immediate inference result viewing.

Expandability: Allows developers to adjust Tensor Arena, input/output handling, and other parameters based on model requirements.

Cross-Platform Support: Uses Miniforge for environment setup to simplify dependency management and toolchain integration.

The toolkit offers a command-line interface for generating AI MCU projects with minimal parameters, eliminating the need for manual complex code writing. It supports a workflow covering model quantization, project building, microcontroller flashing, and result verification. The generated Keil project features a modular co...

You've read this far — sign in to keep reading

Sign in to keep reading.

Forgot password?
OR