Download

Users are strongly recommended to read through the User Manual. It will help clarify some technical details that can help avoid unnecessary errors when dealing with multi-modal data.

Calibration

The calibrated parameters for the sensorsuites are provided in the following yaml files:

Robot Type 1 (for *robot Type1 sequences)

Robot Type 2 (for *robot Type1 sequences)

Single-Agent Scenes

The sequences are captured as rosbags, which are then compressed with bz2 method. User can uncompress the rosbags for less CPU usage at the cost of 3x memory storage. For user convenience, we extracted data from the rosbag, primarily providing color and depth images, lidar points as PCD files, and IMU.

# SeqID Ground Truth Color Image Depth Sensor Image Depth Project Image Livox Imu
1 indoor_long_corridor01 Indoor_Long_Corridor1 Color Depth Sensor Depth Proj Lidar Imu
2 indoor_long_corridor02 Indoor_Long_Corridor2 Color Depth Sensor Depth Proj Lidar Imu
3 indoor_long_corridor03 Indoor_Long_Corridor3 Color Depth Sensor Depth Proj Lidar Imu
4 indoor_long_corridor04 Indoor_Long_Corridor4 Color Depth Sensor Depth Proj Lidar Imu

Multi-Agent Scenes

The multi-agent scenario consists of four trajectories within the long_corridor scene, each containing segments with inter-agent loop closures.

SeqID Ground Truth Color Image Depth Image Livox Imu
Multi-agent_long_corridor Indoor_Long_Corridor1 Color1 Color2 Color3 Color4 Depth1 Depth2 Depth3 Depth4 Lidar1 Lidar2 Lidar3 Lidar4 Imu1 Imu2 Imu3 Imu4