VI-Rig: 5 IMUs + 4 cameras

Demo and values

One MVIS camera + IMU run on recording 5imus_4cams_01: all four cameras and all five IMUs solved together.

The scene MVIS recovered from this dataset, as on the homepage. The values beside it come from this page's run.

Reprojection median0.20 px

cam0–cam1 baseline63.5 mm

cam0 td−0.19 ms

cam0 lever arm193.8 mm

Reprojection is how far, in pixels, the detected corners lie from where the solve puts them. td is the camera–IMU time offset, and the lever arm is the camera–IMU position offset.

Calibration process

MVIS's own player, from this run: the detected board corners, then the rig's path as the solve refines it. Press Play or drag the bar; pick a camera card to look through that camera. We publish no camera images.

Solved rig

The solved positions and orientations, to scale in the IMU frame. Drag sideways to orbit. Tap a sensor to see its values.

3DFrontTopSide+−

Sensor	fx, fy, cx, cy px	Lever mm	log(R) rad	td ms	t_ga ms	Reprojection px
cam0	286.4, 285.6, 415.1, 404.4	[−158.3, −108.7, −26.5]	[+1.5620, −0.0073, −0.0249]	−0.19		0.188
cam1	286.5, 285.4, 414.0, 398.8	[−94.8, −109.5, −27.8]	[+1.5646, −0.0044, −0.0205]	−0.15		0.192
cam2	286.2, 285.1, 416.3, 401.1	[−160.5, −110.4, −49.8]	[+1.5664, +0.0216, −0.0566]	−0.30		0.204
cam3	286.5, 285.5, 412.7, 399.8	[−97.2, −111.1, −53.2]	[+1.5732, +0.0216, −0.0512]	−0.27		0.215
IMU (base)		0	0	0	−0.56 ± 0.17
IMU gx3_25		[−122.0, −46.9, −2.9]	[−2.2372, −2.1947, +0.0186]	−0.43	+0.12 ± 0.20
IMU gx3_35		[−82.3, −47.4, −3.4]	[−2.2388, −2.1989, +0.0128]	−2.01	+0.01 ± 0.21
IMU t265_01		[−145.8, −104.2, −25.9]	[−0.0085, −2.1884, +2.1962]	−7.03	−6.58 ± 0.18
IMU t265_03		[−148.2, −105.5, −51.0]	[+0.0221, −2.1744, +2.1712]	−7.12	−6.59 ± 0.17

Lever arms and rotations are each sensor in the base IMU frame; rotation is log(R) in rad. td is the time offset to the base IMU, t_imu = t_sensor + td. Reprojection is the median per camera (all cameras: median 0.196 px): how well the model fits the corners, not how accurate it is.

Check against a reference

No published reference. No published calibration exists for this rig, so this page has nothing to compare against. What we can show is how much each value moves from one recording to the next. All 7 recordings: Dataset eval →

Config and run

virig_5imus_4cams_cam_imu.yaml

input:
  format: ros1
  path: 5imus_4cams_01.bag
  cam_topics:
  - /t265_01/fisheye1/image_raw
  - /t265_01/fisheye2/image_raw
  - /t265_03/fisheye1/image_raw
  - /t265_03/fisheye2/image_raw
  imu_topics:
  - /mti/sensor/imu_raw
  - /gx3_25/data
  - /gx3_35/imu/data
  - /t265_01/imu
  - /t265_03/imu
  window_mode: motion
  segment_length: 80.0
board:
  board_type: apriltag
  rows: 6
  cols: 6
  tag_size: 0.088
  tag_spacing: 0.3
  tag_family: t36h11
camera_input:
  path: camera_prior.yaml
  cameras:
  - {}
  - factor_hz: 4.0
  - factor_hz: 4.0
  - factor_hz: 4.0
solve:
  estimate_t_ga: true
  max_iterations: 40
  corner_subsample: two_anti
estimate:
  camera:
    intrinsics: true
    td: true
    extrinsics: true
    tr: false
  imu:
    intrinsics: dw_da
    model: calyx
    extrinsics: true
    td: true
  board_points: true
output:
  path: cam_imu.json

shell

mvis cam-imu -c virig_5imus_4cams_cam_imu.yaml

The marked lines are the choices that matter for this rig:

• input.imu_topics: Five IMU topics. The first is the base IMU; the other four are solved against it.
• input.window_mode: MVIS picks the 80 s window with the most motion; this recording is shorter, so all of it is used.
• camera_input.cameras: The three cameras after the first are added to the graph at 4 Hz, which keeps the solve small.
• camera_input.path: The starting camera intrinsics, from camera-only solves of all 7 recordings.
• solve.estimate_t_ga: Estimates t_ga, the time offset between the gyroscope and the accelerometer. The T265 IMUs need it most.
• estimate.board_points: The printed board is not quite flat, so its shape is estimated too.
• estimate.camera.intrinsics: Camera intrinsics stay free in the camera + IMU solve.

This is the config of the run on this page, with every path cut to its file name. Point the paths at your copy.

The solve took about 7 min on one desktop. It writes cam_imu.json, a calibration YAML and a report beside the config.

MVIS

MVIS main, 2026-11-10

Run date

2026-11-10

Data

VI-Rig datasets →

Config

virig_5imus_4cams_cam_imu.yaml, above

Command

mvis cam-imu -c virig_5imus_4cams_cam_imu.yaml

Dataset

• Recording 5imus_4cams_01, one of the 7 recordings of this rig. It is about 59 s long, shorter than the 80 s window MVIS asks for, so the whole recording is used.
• Cameras: the fisheye pairs of two T265 units, all global shutter, with the equidistant lens model. Each pair is synced inside its unit, but the two units are not synced to each other; their time offsets are solved against the base IMU.
• IMUs: the Xsens MTi is the base IMU; the two MicroStrain units and the two T265 IMUs are solved against it.
• Board: AprilGrid 6×6, tag 88 mm, spacing 0.3, t36h11.
• Solve window: 58.7 s.

Download: VI-Rig datasets → The recordings are listed there, not copied here.

AprilGrid 6 × 6, 660 × 660 mm. Standard pattern, to scale