MIT Researchers Creates A New System to Bring Tiny Deep Learning to IoT Devices
MCUNet designs compact neural networks that deliver unprecedented speed and accuracy for deep learning on IoT devices, despite limited memory and processing power. The technology could facilitate the expansion of the IoT universe while saving energy and improving data security. This video is a live visual wake word demo using MCUNet on a $10 microcontroller. MCUNet with TinyNAS and TinyEngine achieves 12% higher accuracy and 2.5× faster speed compared to MobilenetV1 on TF-Lite Micro. The research paper "MCUNet: Tiny Deep Learning on IoT Devices" will be presented at NeurIPS 2020 as a spotlight talk. The lead author is Ji Lin, a PhD student in Song Han’s lab in MIT’s Department of Electrical Engineering and Computer Science. Co-authors include Song Han and Yujun Lin of MIT, Wei-Ming Chen of MIT and National University Taiwan, and John Cohn and Chuang Gan of the MIT-IBM Watson AI Lab.