Machine learning, especially deep learning, based algorithms are gaining popularity in IoT edge devices, as they can offer human-level accuracy in many tasks, such as image classification and speech recognition. We have seen increasing interests in developing and deploying neural networks (NNs) on the types of low-power processors found in always-on IoT Edge systems, such as those based on Arm Cortex-M microcontrollers.
In this talk, we first discuss the challenges of deploying neural networks on microcontrollers with limited memory and compute resources and power budgets. We introduce CMSIS-NN, a library of optimized software kernels to enable deployment of neural networks on Arm Cortex-M processors. We also present techniques for NN algorithm exploration to develop lightweight models suitable for resource constrained systems.