VI-Rig — 4 IMUs + 3 cameras, ArUco tags

The same rig as 4 IMUs + 3 cameras, recorded in front of printed ArUco tags instead of one AprilGrid. ArUco tags are square marker codes, like AprilTags, from another family. The tags are printed on several boards set at different angles, so their corners spread in 3D rather than lying on one plane.

• Hub folder: calibench-vi/datasets → virig-4imus-3cams-aruco
• Licence: CC BY-NC 4.0, gated; commercial use on request. See Licence and access.
• Results: none published. This set is not part of the benchmark.

What it is for

• Feature mode: the tag corners serve as image features whose 3D positions are not given. The calibration estimates those positions together with the sensor parameters, as a visual-inertial system does in the field. No board layout is used.
• In MVIS: the MVIS dataset catalog lists this set for camera + IMU calibration in feature mode only. mvis cam-imu 0.1.0 has no feature mode. One-off study scripts in examples/experiments/ of the repository used it (run_board_vs_feature.py, run_feature_ablation_aruco.py); they are kept to reproduce those studies and are not supported.

The rig

The cameras are the same as in 4 IMUs + 3 cameras: a FLIR Blackfly and a RealSense T265 fisheye pair, all global shutter, at 30 Hz. The IMUs differ between two groups of recordings (rates measured from the bags):

Recordings	IMUs and topics
01–10	all four: Xsens MTi-100 400 Hz /imu/data, GX3-25 500 Hz /gx3_25/data, GX3-35 100 Hz /gx3_35/imu/data, T265 IMU 200 Hz /t265/imu
11–17	three: Xsens MTi-100 400 Hz /imu/data, GX3-25 500 Hz /gx3_25/data, and the T265 gyroscope (200 Hz, /t265/gyro/sample) and accelerometer (about 63 Hz, /t265/accel/sample) on separate topics. No GX3-35.

Recordings

• Count: 17 ROS1 bags, virig_4imus_3cams_aurco_01.bag to virig_4imus_3cams_aurco_17.bag. The file names spell “aurco”; we kept them as recorded.
• Length: 102 to 139 s each, 37 minutes in total.
• Size: 5.4 to 7.3 GB each, 117 GB in total.

Download

Accept the terms on the Hub page and set HF_TOKEN to a read token from your Hugging Face settings (or run hf auth login once, after pip install huggingface_hub). Then either:

# with MVIS: checks size and SHA-256, saves to $MVIS_DATASETS/mvis_4imus_3cams/data_arucotags/
mvis datasets list --dataset mvis_4imus_3cams_aruco
mvis datasets download --select mvis_4imus_3cams_aruco/aruco_01   # one recording, 6.6 GB
mvis datasets download --dataset mvis_4imus_3cams_aruco           # all 17, 117 GB

# with the Hugging Face CLI
hf download calibench-vi/datasets --repo-type dataset \
  --include "virig-4imus-3cams-aruco/*" --local-dir ./calibench

Cite

Please cite Calibench and the MVIS paper. The BibTeX is on the overview page.