UT Campus Object Dataset (CODa): Multimodal Robot Perception Data for Urban Navigation

The University of Texas at Austin campus provides the setting for this mobile robot egocentric perception dataset. It contains 8.5 hours of multimodal sensor data, including synchronized 3D point clouds and stereo RGB video, with 54 minutes of ground-truth annotations featuring 1.3 million 3D bounding boxes across 50 semantic classes. Arthur Zhang published the dataset via the Texas Data Repository in March 2024.

Use Cases

• Benchmarking 3D object detection models based on the 1.3 million annotated bounding boxes.
• Training 3D semantic segmentation algorithms using the 196 million points annotated with terrain labels.
• Developing autonomous navigation planning systems based on the multimodal sensor data collected across diverse urban paths.
• Studying perception robustness across lighting and weather conditions based on sequences collected in cloudy, sunny, dark, and rainy scenarios.

Strengths

• Contains over 196 million 3D points annotated with semantic labels for terrain.
• Includes 1.3 million 3D bounding boxes with instance IDs across 50 semantic classes, described as the largest diversity of object classes in available 3D LiDAR datasets for human-centric urban environments.
• Provides viewpoint diversity with data collected along 4 unique paths traversed in both directions and under 4 different weather/lighting conditions.

Limitations

• Column-level documentation is absent; field semantics must be inferred after download.
• Row count is unknown, which may limit suitability assessment.

Provenance

Source

University of Texas at Austin campus, collected by a team of 7 robot operators.

Collection Method

Data was collected by tele-operating a robot along predefined trajectories, recording synchronized sensor messages via ROS middleware, and post-processing into individual files.

Time Range

null

Freshness

Last updated 2024-03 18 05:50:23; freshness should be verified.

Geography

University of Texas at Austin campus.