Release of a multimodal autonomous driving dataset in a simulation environment

The dataset is free to download, and the collecting tool is open source.

The CARLA-ADA Dataset is a comprehensive multimodal autonomous driving dataset collected using our custom data acquisition tool in the CARLA simulator. The dataset encompasses diverse driving scenarios across urban, rural, highway, and suburban environments, providing a rich variety of real-world driving situations.

The dataset features synchronized image and 3D LiDAR point cloud data, capturing five distinct object classes: cars, trucks, pedestrians, bicycles, and buses. With a total size of 35GB, this dataset offers high-quality annotated data suitable for various autonomous driving perception tasks, including object detection, semantic segmentation, and multi-modal fusion research.

Key Features:

• Multi-environment data collection: urban, rural, highway, and suburban scenes
• Multi-modality: camera images and LiDAR point clouds
• Five object classes with careful annotations
• Diverse weather and lighting conditions
• Comprehensive coverage of real-world driving scenarios

This dataset aims to support research and development in autonomous driving perception systems, particularly for deep learning applications requiring diverse and well-annotated training data.