Bird’s Eye View Perception for Autonomous Driving

Abstract

Bird’s eye view (BEV) perception refers to transforming a perspective view into a bird’s eye view and performing perception tasks such as 3D detection, map segmentation, and motion prediction. Due to its inherent advantages in representing 3D space, fusing multi-modal data, facilitating decision making, and aiding in path planning, BEV perception has garnered significant attention from both academia and industry. In this chapter, we present an overview of BEV perception, covering its definition, categorization, benefits, and mathematical foundations. We then provide a comprehensive review of the current state-of-the-art research, datasets, evaluation metrics, and industrial applications. In the end, we outline several existing challenges and present our own conclusions regarding BEV perception.