Glossary
What Is a Stereo Baseline?
Last reviewed: September 2026
Short answer
The stereo baseline is the fixed distance between the two lenses in a calibrated stereo camera pair. It is one of the calibration values — alongside intrinsics and extrinsics — that a stereo rig needs to turn the disparity between its left and right images into geometric depth for every pixel.
01
Why it matters
Depth recovered from a stereo pair is only as trustworthy as the calibration behind it. If the baseline, intrinsics, or extrinsics drift or are only assumed rather than measured, the resulting depth is off by an amount that compounds with distance from the cameras — which is exactly the error a contact-rich manipulation policy cannot absorb.
02
How Firsthand handles it
Firsthand captures stereoscopic video from a head-mounted stereo rig set at a human interpupillary baseline, matching the viewpoint that humanoid and head-mounted robot cameras actually see from. The baseline, along with intrinsics and extrinsics, is measured every session and shipped with the episode, so downstream rectification and depth recovery are exact rather than assumed. Raw frames ship with that calibration for custom re-rectification, or rectified pairs plus disparity can be added to the delivery.
03
What a stereo baseline is measured against
| Calibration value | What it captures | Why it is shipped per episode |
|---|---|---|
| Baseline | Fixed distance between the two lenses | Sets the scale of recovered depth from disparity |
| Intrinsics | Per-camera lens and sensor parameters | Corrects lens distortion before rectification |
| Extrinsics | Relative pose between the two cameras | Aligns the left and right views for triangulation |
| Stereo pair | The two RGB frames themselves | What the baseline and calibration are applied to |
05
Frequently asked questions
- What is a stereo baseline in simple terms?
- It is the distance between the two lenses in a stereo camera pair. That distance, combined with each lens’s calibration, is what lets a system convert the pixel disparity between the left and right frames into a depth value.
- What baseline distance does Firsthand use?
- A baseline close to human interpupillary distance, matching the viewpoint that humanoid and head-mounted robot cameras actually see from.
- Does the baseline change from session to session?
- The rig is fixed, but Firsthand still measures the baseline along with intrinsics and extrinsics every session and ships those values with the episode, so rectification is exact rather than assumed.
- What happens if the baseline calibration is wrong or missing?
- Depth recovered from disparity is only as accurate as the calibration behind it. An unmeasured or drifted baseline produces depth error that grows with distance from the cameras, which is why Firsthand ships it measured per episode rather than assumed from the rig’s spec sheet.
- Is the stereo baseline the same as the stereo pair itself?
- No. The stereo pair is the two RGB frames; the baseline is one of the calibration numbers — the lens spacing — used to turn that pair into depth. See the stereoscopic video glossary entry for the broader concept.