r2b dataset 2023
r2b Dataset 2023
Dataset Overview
The r2b dataset 2023 is a collection of sequences stored in rosbags, a ROS 2 native format used for message data including time synchronized sensor captures from a robot. The dataset can be loaded and played back into an application running a ROS graph of nodes for testing and performance benchmarking of image processing, perception and localization. Each sequence consists of a single sequence in a scene, providing a small diversity of data across the collection of sequences in this dataset.
Several sequences contain sensor multi-second captures including LIDAR, stereo camera, and IMU data from HAWK stereo camera, RealSense D455, and XT32 with highly precise time synchronization. Other sequences contain data recorded from a simulated scene in Isaac Sim. The r2b_hope sequence is distinct as it contains D415 RGB data.
| Sequence | Visual | Description |
|---|---|---|
r2b_lounge |
 |
Lounge sequence containing couch, table, chairs, and staircase with natural planted background wall. |
r2b_storage |
 |
Storage sequence including person, AprilTag, shoe, shelving, boxes, pallets, with moving obstacles. |
r2b_hallway |
 |
Hallway sequence with walking persons, low to no feature not-perpendicular walls, specular highlights, and reflections. |
r2b_datacenter |
 |
Datacenter sequence with tall vertical corridor, repetitive low-feature surfaces, and little color. |
r2b_cafe |
 |
Café sequence including table, chairs, stools, reflective flooring, dark reflective glass walls, specular highlights, low wall features, and vibration from floor surface. |
r2b_hope |
 |
Images from scene_0005 of the HOPE dataset for 6-DoF pose estimation |
r2b_hideaway |
 |
Hideaway sequence including table, chairs, seated and moving persons, specular highlights, low wall features, and vibration from floor surface. |
r2b_mezzanine |
 |
Mezzanine sequence including staircase, railings, table, chairs, highlights, low wall features, and vibration from floor surface. |
Documentation
Topics Provided
The sequences contained in this dataset contain the following topics:
| Sequence | Topics Provided | Message Type |
|---|---|---|
r2b_lounge r2b_storage r2b_hallway r2b_datacenter r2b_cafe r2b_hideaway r2b_mezzanine |
d455_1_depth_camera_info d455_1_depth_image d455_1_imu d455_1_left_ir_camera_info d455_1_left_ir_image d455_1_rgb_camera_info d455_1_rgb_image d455_1_right_ir_camera_info d455_1_right_ir_image hawk_0_left_rgb_camera_info hawk_0_left_rgb_image hawk_0_right_rgb_camera_info hawk_0_right_rgb_image pandar_xt_32_0_lidar |
sensor_msgs/msg/CameraInfo sensor_msgs/msg/Image sensor_msgs/msg/Imu sensor_msgs/msg/CameraInfo sensor_msgs/msg/Image sensor_msgs/msg/CameraInfo sensor_msgs/msg/Image sensor_msgs/msg/CameraInfo sensor_msgs/msg/Image sensor_msgs/msg/CameraInfo sensor_msgs/msg/Image sensor_msgs/msg/CameraInfo sensor_msgs/msg/Image sensor_msgs/msg/PointCloud2 |
r2b_hope |
camera_info image |
sensor_msgs/msg/CameraInfo sensor_msgs/msg/Image |
MD5 Checksums
A MD5 checksum of each rosbag is provided to verify the dataset sequence used in testing or benchmarking matches, to detect inconsistencies in input data, and to provide a method to confirm that the same input dataset sequence is used when comparing independent results.
| Sequence | Checksum | Size |
|---|---|---|
r2b_lounge |
720ce30055ee3d49e760f384ad402a03 | 3.9G |
r2b_storage |
5e8f11201fe10dbac7307a8628553e94 | 2.9G |
r2b_hallway |
7cc6b54751d37a78893d5bea43b3fbf5 | 1.3G |
r2b_datacenter |
188167fd025d5922940fb9531f835691 | 1.7G |
r2b_cafe |
8ef0e6b552cb19b7486cc747f3f3fa43 | 1.2G |
r2b_hope |
22260103ef97ce5cf563e2c6ee656a4e | 30M |
r2b_hideaway |
96a3c2b8007173f9da57e848b4fcee8c | 1.8G |
r2b_mezzanine |
c84d7420c5524d882f1019d350232fb1 | 2.0G |
Getting Started/Example
Download the dataset to your local system, and playback the rosbag for test or benchmarking.
ros2 bag play path/to/r2b_sequence
For more information about using this dataset for benchmarking ROS 2 graphs of nodes, see ros2_benchmark.
License
This dataset is licensed under Creative Commons - Attribution 4.0 International.
Ethical Considerations
This dataset consists of captures in NVIDIA’s facilities in the United States. Persons in the datasets have consented to use of their images.
This dataset is not sufficiently diverse to be representative of the public in general.