In this dataset we collected a series small-scale indoor table AR visual-inertial datasets ranging from 1-2 minutes in length. A Intel RealSense Depth Camera D455 is equipped with groundtruth Optitrack markers and provides color monocular image at 30Hz and a 400Hz IMU inertial feed from its Bosch BMI055 IMU. Additionally, the depth projected into the color image by the realsense driver is also recorded (untested). Groundtruth for the IMU of each platform is found through the vicon2gt utility package developed by our group.
If you are interested in running visual-inertial odometry on this dataset, please checkout the ov_plane project along with OpenVINS. Both can run on this dataset, with the former supporting estimation of enviromential planes to improve performance and enable augumented experiences. Please check it out!
/d455/color/image_raw : sensor_msgs/Image (rgb image 30hz)
/d455/imu : sensor_msgs/Imu (imu data 400hz)
/d455/aligned_depth_to_color/image_raw : sensor_msgs/Image (depth 30hz)
/d455/rigidbody : geometry_msgs/PoseStamped (optitrack)
Each dataset has the system starting and ending at around the same location and starting from stationary.
The groundtruth for trajectory evaluation can be found in the groundtruth/ folder.
We provide the original bags along with compressed bags for slower internets.
To uncompress the bags one can run rosbag decompress <bag_name> after ensuring there is enough space to create a copy of the bag.
| Bag Raw | Bag Compressed | Size (GB) | Length (m) | Video (x10) | |
|---|---|---|---|---|---|
| table1 | rosbag | rosbag | 4.77 | 55.6 | link |
| table2 | rosbag | rosbag | 5.54 | 43.6 | link |
| table3 | rosbag | rosbag | 13.19 | 87.9 | link |
| table4 | rosbag | rosbag | 11.49 | 91.4 | link |
| table5 | rosbag | rosbag | 11.66 | 75.5 | link |
| table6 | rosbag | rosbag | 5.26 | 49.9 | link |
| table7 | rosbag | rosbag | 9.02 | 63.0 | link |
| table8 | rosbag | rosbag | 16.01 | 125.0 | link |
A single dataset was used for both static and dynamic calibration of the sensors using Kalibr with the april_6x6_80x80cm.yaml calibration board. Additionally, a 20 hour static IMU dataset was recorded to recover the IMU intrinsic noise properties (random walk and biases). The parameters were found using the allan_variance_ros toolbox. Result files can be found in the calibration/ folder.
| ROS Bag | Results | |
|---|---|---|
| color1 | download | imu-camchain |
| color2 | download | imu-camchain |
| static | download | plots |
This dataset was collected by the Robot Perception and Navigation Group (RPNG) at the University of Delaware. If you have any issues with the data please open an issue on our github page with relevant implementation details and references. For researchers that have leveraged or compared to this work, add a footnote to this repository, or use the following citation:
@Conference{Chen2023ICRA,
Title = {Monocular Visual-Inertial Odometry with Planar Regularities},
Author = {Chuchu Chen and Patrick Geneva and Yuxiang Peng and Woosik Lee and Guoquan Huang},
Booktitle = {Proc. of the IEEE International Conference on Robotics and Automation},
Year = {2023},
Address = {London, UK},
Url = {\url{https://github.com/rpng/ov_plane}}
}