SynthFlood-Car: Synthetic and Refined Dataset for Vehicle Flood-Level Detection (Dataset)

SynthFlood-Car: A Synthetic and Annotated Dataset for Vehicle Flood Level Detection UrbanFloodSynth is a training dataset based on both synthetic and real-world data, designed to detect vehicle flood levels in urban flooding environments. This dataset was constructed using a 3D simulation-based data generation pipeline to address the challenges of collecting real-world flood data and the high costs associated with labeling. Additionally, real-world data was utilized in conjunction with a Human-in-the-Loop (HITL)-based labeling refinement process to improve the model’s generalization performance and robustness. This dataset includes various urban environments, vehicle types, and flood water level conditions, and can be used for object detection training to detect vehicles at different flood stages. It supports a wide range of real-world applications, including disaster response, smart city monitoring, and flood risk analysis.