Hardware-Algorithm Co-Design Enabling Efficient Event-based Object Detection

Event-based cameras are a promising alternative to traditional optical cameras for real time computer vision systems. They offer low power, high temporal resolution, and high dynamic range, making them an ideal candidate for resource-constrained edge computing. In this work, we evaluate system level designs using event-based cameras for computer vision on the edge. We present a new technique to improve the object detection performance of event-based cameras, and demonstrate a $1.88 \times$ to $2.17 \times$ improvement in energy efficiency.