In recent years, edge-vision monitoring systems for applications such as smart animal husbandry have faced strict tripartite constraints: maintaining input resolution under extremely limited transmission bandwidth and strict power budgets. Conventional dense convolutional neural…
In deep learning, efficiency gets more and more important to compensate for the ongoing growth in model sizes and applications. Neuromorphic hardware has long been advocated as an upcoming alternative to deep networks, taking inspiration from the brain for achieving unprecedented…