Guest
tinyML Asia Partner Session SPECK– A Low power, low latency neuromorphic visual solution in a single chip Yannan XING , Senior Algorithms Engineer, SynSense While DNNs with accelerators have demonstrated remarkable performance in visual tasks, this gain comes with undesired costs — e.g. redundant computation and long processing latency. This is especially undesirable in the field of IoTs and other edge computing applications. Neuromorphic Computing takes inspiration from efficient computation in biological neural networks, to build low-power ML applications. By using special-purpose computing architectures, neuromorphic computing introduces significant benefits for low-power and low-latency processing. SynSense has developed a commercial neuromorphic smart vision sensor —SPECK — which integrates an event-based Dynamic Vision Sensor (DVS) and Neuromorphic Computing cores into an single-die SoC. SPECK provides end-to-end ultra low power, real-time solutions for application scenari
