Talk2BEV: Language-enhanced Bird’s-eye View Maps for Autonomous Driving

Vikrant Dewangan∗1    Tushar Choudhary∗1    Shivam Chandhok∗2    Shubham Priyadarshan1    Anushka Jain1    Arun K. Singh3    Siddharth Srivastava4    Krishna Murthy Jatavallabhula†5    K. Madhava Krishna†1   

1 Robotics Research Center, IIIT Hyderabad, India    2 University of British Columbia    3 University of Tartu    4 TensorTour Inc    5 MIT Computer Science & Artificial Intelligence Laboratory   



This work introduces Talk2BEV, a large vision-language model (LVLM)1 interface for bird’s-eye view (BEV) maps in autonomous driving contexts. While existing perception systems for autonomous driving scenarios have largely focused on a pre-defined (closed) set of object categories and driving scenarios, Talk2BEV blends recent advances in general-purpose language and vision models with BEV-structured map representations, eliminating the need for task-specific models. This enables a single system to cater to a variety of autonomous driving tasks encompassing visual and spatial reasoning, predicting the intents of traffic actors, and decision-making based on visual cues. We extensively evaluate Talk2BEV on a large number of scene understanding tasks that rely on both the ability to interpret free-form natural language queries, and in grounding these queries to the visual context embedded into the language- enhanced BEV map. To enable further research in LVLMs for autonomous driving scenarios, we develop and release Talk2BEV-Bench, a benchmark encompassing 1000 human-annotated BEV scenarios, with more than 20,000 questions and ground-truth responses from the NuScenes dataset.