An Initial Study of Bird's-Eye View Generation for Autonomous Vehicles using Cross-View Transformers

TL;DR

This paper explores using Cross-View Transformers to generate Bird's-Eye View maps from camera images for autonomous driving, focusing on generalization and robustness in urban environments.

Contribution

It introduces a novel application of Cross-View Transformers for BEV map generation and evaluates their performance across different camera layouts and unseen towns.

Findings

Four-camera CVT with L1 loss performs best in new towns.

CVT demonstrates promising generalization to unseen urban areas.

L1 loss yields more robust BEV maps than focal loss.

Abstract

Bird's-Eye View (BEV) maps provide a structured, top-down abstraction that is crucial for autonomous-driving perception. In this work, we employ Cross-View Transformers (CVT) for learning to map camera images to three BEV's channels - road, lane markings, and planned trajectory - using a realistic simulator for urban driving. Our study examines generalization to unseen towns, the effect of different camera layouts, and two loss formulations (focal and L1). Using training data from only a town, a four-camera CVT trained with the L1 loss delivers the most robust test performance, evaluated in a new town. Overall, our results underscore CVT's promise for mapping camera inputs to reasonably accurate BEV maps.