ParkingScenes dataset released for end-to-end autonomous parking in simulated environments with structured trajectories and multimodal sensor data.

ParkingScenes is a multimodal dataset built on the CARLA simulator, containing 704 structured episodes and approximately 105000 frames across 16 reverse-in and 6 parallel parking scenarios, each executed under two pedestrian conditions (present and absent) and repeated 16 times.

Each frame includes synchronized data from four RGB cameras, four depth sensors, vehicle motion states, and Bird's-Eye View (BEV) representations; trajectories are generated by a Hybrid A* planner and Model Predictive Controller (MPC) for accurate supervision signals.

Models trained on ParkingScenes showed significant performance improvements compared with those trained on unstructured, manually collected simulation data under identical conditions, demonstrating the effectiveness of structured supervision for parking policy learning.