Dataset and Scenario-based Evaluation Framework

The evaluation is conducted using a large-scale synthetic dataset comprising approximately 700,000 synchronised multi-modal samples generated within the CARLA autonomous driving simulator. The dataset is designed to support comprehensive scenario-based evaluation across diverse driving conditions, including variations in traffic density, weather conditions, lighting, and road topology. Each sample includes temporally aligned sensor modalities such as RGB images, depth maps, semantic segmentation, LiDAR point clouds, and vehicle state information, enabling robust multi-modal learning and benchmarking.