"Low-cost neural network inferencing on the edge with xcore.ai"
Machine Learning Engineer
XMOS recently launched its next generation crossover processor, xcore.ai, featuring a novel vector unit designed for low precision integer and binarized neural network inference. In this session, we introduce the core ideas behind this vector unit, and explain how it deviates from a traditional load-store architecture to enable high throughput when calculating convolutions. We go on to outline the software tools and libraries that enable users to take full advantage of the hardware. Lastly, we demonstrate our optimization and deployment tools based on TensorFlow Lite for microcontrollers, by converting and analyzing a variant of MobileNet.
"Low Power Embedded Gesture Recognition Using Novel Short-Range Radar Sensors"
Department of Information Technology and Electrical Engineering
Human-computer interface (HCI) is an attractive scenario. A promising novel sensing technology is high-frequency short-range Doppler-radar. This talk presents a low-power high-accuracy embedded hand-gesture recognition using low power short-range radar sensors from Acconeer. A 2D Convolutional Neural Network (CNN) using range frequency Doppler features is combined with a Temporal Convolutional Neural Network (TCN) for time sequence prediction. The final algorithm has a model size of only 45723 parameters, yielding a memory footprint of only 91kB. We acquired two datasets containing 11 challenging hand gestures performed by 26 different people containing a total of 20210 gesture instances. The algorithm achieved an accuracy of up to 92% on the 11 hands gestures. Furthermore, we implemented the prediction algorithm on the GAP8 Parallel Ultra-Low-Power processor RISC-V and ARM Cortex-M processors. The hardware-software solution matches the requirement for battery-operated wearable devices.