Low-Cost Neural Network Inferencing on the Edge With Xcore.AI
Explore low-cost neural network inferencing on edge devices with xcore.ai in this tinyML Talks webcast. Discover XMOS's next-generation crossover processor featuring a novel vector unit designed for low precision integer and binarized neural network inference. Learn about the core ideas behind this vector unit and how it differs from traditional load-store architecture to enable high-throughput convolution calculations. Gain insights into the software tools and libraries that allow users to maximize the hardware's potential. Witness a demonstration of optimization and deployment tools based on TensorFlow Lite for microcontrollers, including the conversion and analysis of a MobileNet variant. The talk covers topics such as the vector unit infrastructure, optimized device demonstrations, and future developments in the field.