Architecture Design for Pedestrian Detection Based on Memory Grid Occupancy and Data Reuse

Pedestrian Detection, as an integral component of autonomous driving systems, is directly related to the vehicle safety. However, trapped by the current computing paradigm based on Von Neumann architecture, the traditional design of Pedestrian Detection system faces certain dilemmas in terms of energy consumption and computational efficiency. In this paper, an approach based on Memory Grid Occupancy (MGO) and data reuse is proposed to optimize the computational process of the Pedestrian Detection system. The results of the related work indicate that the proposed optimization design in hardware can reduce the data read/write during the execution of the system.