ChatBEV: Empowering Traffic Scene Understanding and Simulation via Vision-Language Model
ChatBEV-QA is introduced, a large-scale BEV VQA benchmark of 137K+ QA pairs, designed to evaluate global scene understanding, vehicle-lane interactions, and vehicle interactions within complex traffic environments.
Qing-Yao Xu, Ya Zhang, Yan-Feng Wang et al.
· 6 citations