CARLA-to-nuScenes

Data Format Conversion Tool

Convert CARLA simulation data to nuScenes format for seamless integration with autonomous driving frameworks

Overview

We provide a powerful data conversion tool that can transform data collected from the CARLA simulator into nuScenes format. This tool supports multiple data modalities including LiDAR point clouds, RGB images, GPS/IMU data, and more, enabling researchers to conduct autonomous driving research under a unified data format.

Currently, CARLA-collected data supports the OpenCOOD format, which is a data format specifically designed for cooperative perception. Through our conversion tool, you can easily convert this data to the widely-used nuScenes format, allowing you to leverage existing models and tools in the nuScenes ecosystem.

Key Features

Multi-Format Support

Support complete conversion from OpenCOOD format to nuScenes format while maintaining data integrity and annotation accuracy.

Multi-Modal Data

Convert LiDAR point clouds, RGB images, radar data, GPS/IMU sensor data, and corresponding annotation information.

Cooperative Scenarios

Specifically optimized for multi-vehicle cooperative scenario data conversion, supporting coordinate frame transformations and temporal synchronization between vehicles.

Easy Integration

Provides simple APIs and configuration files that can be quickly integrated into existing data processing pipelines.

Supported Data Types

Sensor Data

  • LiDAR Point Clouds (.pcd/.bin)
  • RGB Camera Images (.png/.jpg)
  • Depth Images
  • Semantic Segmentation Images
  • Radar Data

Annotation Data

  • 3D Bounding Boxes
  • 2D Bounding Boxes
  • Object Trajectories
  • Semantic Labels
  • Instance Segmentation

Metadata

  • GPS/IMU Data
  • Vehicle Poses
  • Sensor Calibration
  • Timestamps
  • Scene Descriptions

Usage Guide

We provide data conversion services to transform your CARLA data into nuScenes format. Contact us with your dataset details and conversion requirements.

Morui Zhu
moruizhu@my.unt.edu
Qing Yang
Qing.Yang@unt.edu

Please include in your email:

  • Dataset description (size, scenarios, sensors)
  • Current data format and desired specifications
  • Timeline and any special requirements