ContextVLM: Zero-Shot and Few-Shot Context Understanding for Autonomous Driving using Vision Language Models | Synapse