RealSense D435i
Depth camera: global-shutter infrared stereo, rolling-shutter RGB and a 6-axis IMU

- Cameras
- 640 × 480 here: RGBMeasured
- Shutter
- Rolling (RGB)Maker ↗︎
- Field of view
- Calibrated 56° × 43°Measured
69° × 42° RGB, 87° × 58° depthMaker ↗︎ - Baseline
- 50 mm, infrared pair (not used here)Maker ↗︎
- IMU
- BMI055, 200 HzMeasured
- Output
- USB-C, EuRoC mav0Maker ↗︎
Maker values checked 2026-11-14 · 3 recordings, 67.7 to 80 s
Demo and values
Section titled “Demo and values”Depth camera from RealSense with a 50 mm global-shutter infrared stereo pair, a rolling-shutter RGB camera and a 6-axis IMU. Three hand-held recordings in front of an AprilGrid, each solved as RGB + IMU with the rolling-shutter readout estimated.
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.
The three recordings agree on the time offset (6.00 to 6.12 ms), the readout time (32.5 ms) and the lever arm across the image (within 0.4 and 1.4 mm), but not along the optical axis: z is 14.7 to 20.4 mm, against 11.3 mm in the RealSense SDK's design values. All three recordings turn the camera little about the two axes across the image, which is what makes that distance observable. Holding z at the SDK value fits nearly as well (within about 2 σ), so the SDK value is the safer choice for now.
Calibration process
Section titled “Calibration process”The path of the rig around the board, as MVIS solved it. To explore it:
- Press Play or drag the bar to follow the path.
- Use the view buttons to change the view.
- Pick a camera card to put that camera on the frustum, the pyramid that shows its view.
Each card draws the board from the solved pose, with the corners MVIS detected. We publish no camera images.
Solved rig
Section titled “Solved rig”The solved rig in 3D, to scale. Tap a sensor to see its values. In full screen you can rotate and zoom. Grid squares are 20 mm.
To scale, from above and the front
| Camera | Lever x, y, z (mm) | Distance (mm) | log(R) (rad) | td (ms) | Field of view |
|---|---|---|---|---|---|
| cam0 | −20.8, +8.0, +20.4 | 30.2 | [+0.003, −0.005, −0.001] | +6.00 | 56° × 43° |
| Camera | Lever x, y, z (mm) | Distance (mm) | log(R) (rad) | td (ms) | Field of view |
|---|---|---|---|---|---|
| cam0 | −20.5, +8.5, +19.0 | 29.2 | [+0.004, −0.008, −0.000] | +6.03 | 56° × 43° |
| Camera | Lever x, y, z (mm) | Distance (mm) | log(R) (rad) | td (ms) | Field of view |
|---|---|---|---|---|---|
| cam0 | −20.8, +7.2, +14.7 | 26.5 | [+0.005, −0.009, −0.002] | +6.12 | 55° × 43° |
Solved by MVIS. Lever arms (camera–IMU position offsets) are in the IMU frame. Rotation is log(R), camera in the IMU frame.
Specifications
Section titled “Specifications”| Published by the maker | |
|---|---|
| Maker | RealSense (formerly Intel RealSense) |
| Shutter | Global on the infrared pair, rolling on the RGB camera |
| Infrared pair | 87° × 58° depth field of view; not used here |
| RGB camera | 2 MP, 69° × 42°, up to 1920×1080 at 30 fps; recorded here at 640×480, 30 Hz, exposure fixed at 3 ms |
| Link | USB-C 3.1 Gen 1; 90 × 25 × 25 mm |
| Measured in these recordings | |
| IMU | Bosch BMI055; gyro and accelerometer at 200 Hz in these recordings, accelerometer resampled onto the gyro stamps |
| Calibrated field of view | cam0 56° × 43° |
Dataset
Section titled “Dataset”| Recordings | 3 recordings, 80, 67.7 and 77 s, hand-held. About 2,000 to 2,400 RGB frames each, IMU at 200 Hz |
| Format | EuRoC mav0: RGB PNG and IMU CSV |
| Board | AprilGrid, 6×6 tags of 88 mm, spacing 0.3 |
| Exposure | Fixed, 3 ms; gain fixed after a short warm-up; white balance fixed |
| Data | Not hosted |
Detailed results
Section titled “Detailed results”Lens Calibrated field of view
Calibrated field of view, horizontal × vertical. cam0: 56° × 43°, corners reach 33° from the axis
Calibrated field of view, horizontal × vertical. cam0: 56° × 43°, corners reach 34° from the axis
Calibrated field of view, horizontal × vertical. cam0: 55° × 43°, corners reach 33° from the axis
cam0
We read these from the calibrated lens fit at the image edges, through the principal point (where the optical axis meets the image). ~ marks an axis past where the corners reached, so the fit is extrapolated there. Past the fit means an edge lies beyond where the fit holds, so we give no value.
IMU Gyro–accel time offset t_ga −3.87 msGyro–accel time offset t_ga −3.81 msGyro–accel time offset t_ga −3.43 ms
Each bar shows a scale as its deviation from 1. IMU model: calyx. t_ga, the time offset between the gyroscope and accelerometer channels: −3.87 ± 0.53 ms.
Each bar shows a scale as its deviation from 1. IMU model: calyx. t_ga, the time offset between the gyroscope and accelerometer channels: −3.81 ± 0.46 ms.
Each bar shows a scale as its deviation from 1. IMU model: calyx. t_ga, the time offset between the gyroscope and accelerometer channels: −3.43 ± 0.72 ms.
Timing cam0 td, first row +6.00 mscam0 td, first row +6.03 mscam0 td, first row +6.12 ms
The time offsets, to scale. td is the camera–IMU time offset, with t_imu = t_cam + td. For a rolling-shutter camera td is the time of the first image row; the middle row is td + readout / 2, which is what a global-shutter calibration reports.
Board shape peak to peak 5.8 mmpeak to peak 6.0 mmpeak to peak 6.0 mm
The calibration board isn't perfectly flat, so this solve estimated the height of every corner off the flat board along with the calibration. The board's mean plane and tilt belong to its pose, so a flat board shows only noise here.
Full results Every solved parameter
Rotation is log(R) in rad, camera in the IMU frame.
| Parameter | MVIS |
|---|---|
| cam0 | |
| fx px | 599.37 |
| fy px | 598.02 |
| cx px | 333.53 |
| cy px | 249.02 |
| k1 | 0.10194 |
| k2 | −0.20872 |
| k3 | 0.00257 |
| k4 | −0.00001 |
| Lever x mm | −20.76 |
| Lever y mm | +8.01 |
| Lever z mm | +20.40 |
| log(R) rad | [+0.0026, −0.0054, −0.0010] |
| td ms | +6.003 |
| Reprojection px | 0.124 |
| Readout ms | +32.55 |
| IMU | |
| t_ga ms | −3.87 ± 0.53 |
| Parameter | MVIS |
|---|---|
| cam0 | |
| fx px | 599.57 |
| fy px | 598.27 |
| cx px | 332.83 |
| cy px | 248.45 |
| k1 | 0.10413 |
| k2 | −0.21446 |
| k3 | 0.00223 |
| k4 | −0.00034 |
| Lever x mm | −20.46 |
| Lever y mm | +8.54 |
| Lever z mm | +19.04 |
| log(R) rad | [+0.0038, −0.0078, −0.0003] |
| td ms | +6.032 |
| Reprojection px | 0.132 |
| Readout ms | +32.56 |
| IMU | |
| t_ga ms | −3.81 ± 0.46 |
| Parameter | MVIS |
|---|---|
| cam0 | |
| fx px | 601.89 |
| fy px | 600.52 |
| cx px | 332.88 |
| cy px | 248.97 |
| k1 | 0.10382 |
| k2 | −0.21002 |
| k3 | 0.00261 |
| k4 | −0.00041 |
| Lever x mm | −20.84 |
| Lever y mm | +7.17 |
| Lever z mm | +14.72 |
| log(R) rad | [+0.0054, −0.0092, −0.0024] |
| td ms | +6.122 |
| Reprojection px | 0.125 |
| Readout ms | +32.53 |
| IMU | |
| t_ga ms | −3.43 ± 0.72 |
Limits
Section titled “Limits”- The three recordings agree on the time offset (6.00 to 6.12 ms), the readout time (32.5 ms) and the lever arm across the image (within 0.4 and 1.4 mm), but not along the optical axis: z is 14.7 to 20.4 mm, against 11.3 mm in the RealSense SDK's design values. All three recordings turn the camera little about the two axes across the image, which is what makes that distance observable. Holding z at the SDK value fits nearly as well (within about 2 σ), so the SDK value is the safer choice for now.
- The tag size, 88 mm, is the stated size of the printed board, and it scales every lever arm.
- The board is not flat: its corners lie up to about 6 mm out of plane, the same shape in every recording.
- The camera stores no IMU correction, so the IMU scale and bias values are those of the raw sensor.
How it was run
Section titled “How it was run”- RGB camera with the IMU in one solve per recording, as a rolling-shutter camera with the readout time estimated. Pinhole camera with radial-tangential distortion, seeded from a single-camera solve of the same recording and left free; the camera time offset, IMU scale and misalignment and the gyro-to-accel offset t_ga estimated. The board shape is estimated per corner. IMU noise is the MVIS default, since no measured values exist for this unit. Solved 2026-11-14.