Glossary

What Is Stereoscopic Video?

Video captured by two horizontally offset cameras at once, giving each frame recoverable depth.

Last reviewed: September 2026

Short answer

Stereoscopic video is footage captured by two horizontally offset cameras at once, so each frame carries the same disparity cues human binocular vision uses for depth. Paired with per-camera calibration, it lets a downstream system recover 3D structure directly from the images rather than only from a separate depth sensor.

01

Why it matters

Monocular video leaves depth ambiguous — a model has to infer scale and distance from cues like size and occlusion, which breaks down on novel objects. Stereo pairs give geometric depth directly from disparity, which matters most for contact-rich manipulation where millimeters of reach and clearance decide success. See the fuller comparison in the stereo vs. monocular guide below.

02

How Firsthand handles it

Firsthand captures stereoscopic video from a head-mounted stereo rig at a human interpupillary baseline, hardware-synced on one clock with depth, IMU, and hand pose. Intrinsics, extrinsics, and the stereo baseline are measured per session and shipped with the episode, and 3D hand joints, object poses, and contacts are annotated against the stereo geometry rather than guessed from a single frame. Raw frames ship with calibration for custom re-rectification, or rectified pairs plus disparity can be added to the delivery.

03

Stereoscopic vs. monocular capture, for training data

PropertyMonocularStereoscopic
Depth signalInferred (size, occlusion, motion)Recovered from disparity, geometric
Sensitivity to novel objectsDepth estimate degradesDepth still measured directly
3D hand/object poseEstimated from a single viewTriangulated against stereo geometry
Calibration requiredNonePer-session intrinsics, extrinsics, baseline
Firsthand deliveryHead camera + wrist camerasHead-mounted stereo pair, raw or rectified

04

Frequently asked questions

Is stereoscopic video the same as depth video?
No. Stereoscopic video is the pair of RGB frames itself; depth is a value that can be computed from that pair (or measured by a separate depth sensor). Firsthand can ship the stereo pair, a computed disparity/depth map, or both on one clock.
Why use stereo instead of a single camera plus a depth sensor?
Stereo depth is pixel-aligned with the RGB view by construction and works outdoors and in conditions where active depth sensors struggle. Many programs capture both and sync them on one clock.
Do you deliver raw or rectified stereo pairs?
Both, if you want. Raw frames ship with the per-session calibration so you can re-rectify yourself, and rectified pairs plus disparity can be added to the delivery.
What lens spacing does Firsthand use?
A baseline close to human interpupillary distance, matching the viewpoint that humanoid and head-mounted robot cameras actually see from.
Can stereo capture be combined with wrist cameras and hand pose?
Yes. Head-mounted stereo, wrist views, depth, IMU, and 3D hand pose all trigger on the same hardware clock in one episode.