Until recently, AI on tiny microcontrollers was deemed impossible. Now, thanks to tools like Mbed and TensorFlow Lite for Microcontrollers, it is not only possible, it is easy and within the reach of every open-source software developer, maker, and start-up.
Soon, more optimized low-level kernels will be available as part of the CMSIS-NN open-source project. These kernels allow developers to leverage Single Instruction Multiple Data (SIMD) instructions and receive an uplift in performance. SIMD instructions are available in Arm Cortex-M4, Cortex-M7, Cortex-M33, and Cortex-M35P processors.
Now that you have implemented your first machine learning application on a microcontroller, it is time to get creative.
To keep learning about ML with Arm and TensorFlow, here are some additional resources:
- View another in-depth end-to-end TensorFlow from training to deployment guide
- Read a white paper on CMSIS-NN: Machine learning on Arm Cortex-M Microcontrollers
- Learn how to use the OpenMV camera for ML applications
- Explore Arm NN for ML on other Arm processors and GPUs
Share your projects using the hashtag #TinyML, and tag @Arm.