Evaluation and Fusion of Vision-Language and Computer Vision Models for On-road Scenario Extraction in Autonomous Vehicles