Egocentric Video Data for Office & Workplace

Quick facts

Resolution

1080p @ 30fps baseline; 2160p for fine document- and screen-reading detail

Field of view

≥120° horizontal — keeps both hands, the desk surface and a nearby colleague in frame at once

Mount

Head-mounted glasses or cap rig — never chest-mount or handheld, which lose the gaze-directed target during desk and screen work

Sensors

RGB (baseline), IMU (head, optional wrist), Gaze (optional — attention on a screen, document or person), Depth (optional — indoor layout and navigation)

Labels

Frame-aligned activity segments (desk work, meeting, transit, handling, service interaction); Person and interaction labels (who is present, turn-taking, hand-offs); Object and device states (screen on/off, document open, equipment in use)

Volume

40–200 accepted hours per workplace/task program; pilot batch in days

What office & workplace egocentric data captures

Office egocentric video data is first-person, head-mounted footage of workplace tasks — desk and device work, meetings and collaboration, moving through indoor space, and handling documents and equipment from the worker's own point of view. The nearest open indoor corpora (EgoBody, HoloAssist, Aria Digital Twin, Aria Everyday Activities) are all non-commercial or research-only, so consented custom capture is the only clean route to commercially-usable workplace training data.

The capture settings this covers:

• Desk and device work: typing, reading and annotating documents, and switching attention between monitor, laptop and phone from the wearer's own viewpoint.
• Meetings and collaboration: two or more colleagues at a whiteboard or table — gesturing, passing documents, and pointing at a shared screen.
• Moving through indoor space: walking corridors, taking stairs and lifts, opening doors, and navigating between desks, meeting rooms and shared areas.
• Handling office equipment: operating a printer or copier, plugging in cables and adapters, sorting mail, and using a kitchenette or coffee machine.
• Front-desk and service interactions: greeting visitors, signing for deliveries, and reception or help-desk exchanges with people in frame.
• Facilities and light maintenance: setting up AV kit before a meeting, swapping a toner cartridge, or resetting network hardware in a comms cupboard.

Why robotics and AI labs need office & workplace data

Apple's EgoDex pretrains dexterous manipulation on large-scale egocentric human video, the first-person desk-and-device footage an office-assistant policy learns from. ^[1]

EgoLive shows large-scale egocentric human demonstrations of real-world tasks lifting manipulation policies, so consented head-mounted footage of office work directly improves the assistant skill it targets. ^[2]

AoE frames scalable, low-cost collection of egocentric human video of everyday tasks as an answer to the data scarcity an office-workplace corpus is built to close. ^[3]

EgoScale reports dexterous-manipulation performance scaling with the volume and diversity of egocentric human data, which makes broad environment coverage — office and indoor included — a measurable lever, not a nice-to-have. ^[4]

Capture and delivery spec

Every office & workplace capture program runs to an explicit spec so the footage is training-ready on delivery rather than after a re-shoot. The baseline below is tuned per program; sensors, labels, and volume scale with the buyer's model.

Open office & workplace datasets

The 5 open corpora most relevant to office & workplace are compared below on scale, sensors, license, commercial use, and the gap each leaves for a buyer. None of them are cleanly licensed for commercial model training — which is the whole reason custom capture exists.

Open datasets vs Truelabel custom capture

Licensing reality: there is no commercially-usable open office corpus. EgoBody is non-commercial, Aria Digital Twin and Aria Everyday Activities are non-commercial under the Aria research license, and even HoloAssist — the most permissive of the set — is a research license you have to verify before shipping. Custom capture attaches a signed wearer consent and workplace release to every clip, so the footage is rights-clean by construction.

Instrumented-room and staged-task gap: the open indoor corpora are either two instrumented rooms with laboratory ground truth (Aria Digital Twin) or scripted instructor-follower repair sessions (HoloAssist). Neither is the messy variety of a real working office — the corridors, hot-desks, reception, and impromptu meetings your model has to handle. Custom capture is shot in an actual workplace, to your task list.

Multi-person and consent are the hard part in an office. Workplace footage is full of colleagues and visitors, so provenance is not optional and colleague/bystander consent — not license alone — is the real gate. Custom capture defines the release chain up front (wearer, colleagues, and a facility release), and can blur or exclude anyone who has not signed, which the open corpora do not document per clip.

Taxonomy and exclusivity: open corpora freeze someone else's activity and interaction schema, and every competitor has already trained on them. Custom capture lets you define the desk/meeting/transit/service taxonomy and person-interaction labels your model consumes, delivered in your episode format and exclusive to you.

Office & workplace: by the numbers

The figures below are specific to office & workplace egocentric data and anchor the comparisons above.

• HoloAssist: ~166 h of two-person assistive/repair sessions on HoloLens 2 — the office-adjacent workplace-assistant reference corpus
• Aria Everyday Activities: ~7.3 h — the entire open 'everyday indoor' Aria corpus is smaller than a single office pilot batch
• Aria Digital Twin: instrumented apartment + office rooms with GT object/human poses, depth and segmentation (non-commercial)
• EgoBody: 3D body shape + motion ground truth for pairs of interacting people, captured from head-mounted devices
• EgoExoLearn: ~120 h of paired ego-exo procedural task-following across daily-life and lab settings
• "indoor scene dataset" 10/mo secondary; "office egocentric video dataset" 0/mo — a pre-volume land-grab (DataForSEO 2026-07-06)

How Truelabel captures office & workplace data

Truelabel runs office & workplace programs on a network of 20,000+ consented collectors across nine countries, capturing to your brief on a head-mounted rig. Every clip passes per-clip machine QA — head-mount stability, field of view, and hands-in-frame coverage — and ships with a signed wearer consent artifact and provenance manifest. A calibration pilot returns its first batch in days, then accepted batches scale to 40–200 accepted hours per workplace/task program; pilot batch in days, delivered as H.265 clips + per-clip JSON metadata (activities, people, device states, navigation), Hugging Face-streamable; consent and workplace-release artifacts shipped per clip. Go deeper via what egocentric data is, egocentric data licensing, the physical AI data marketplace, VLA training data, humanoid robot training data, and industrial egocentric video sourcing.

Related pages

Use these to move from category-level context into specific task, dataset, format, and comparison detail.

External references and source context

1. EgoDex: Learning Dexterous Manipulation from Large-Scale Egocentric Video

   EgoDex is a large-scale egocentric video corpus built to pretrain dexterous manipulation from first-person human video.

   arXiv ↩
2. EgoLive: A Large-Scale Egocentric Dataset from Real-World Human Tasks

   EgoLive is a large-scale egocentric dataset of real-world human tasks used to lift manipulation policies.

   arXiv ↩
3. AoE: Always-on Egocentric Human Video Collection for Embodied AI

   AoE frames scalable, low-cost collection of egocentric human video of manual tasks as an answer to embodied-AI data scarcity.

   arXiv ↩
4. EgoScale: Scaling Dexterous Manipulation with Diverse Egocentric Human Data

   EgoScale reports egocentric-data-trained manipulation performance scaling with the volume and diversity of the human data, motivating broad environment coverage.

   arXiv ↩
5. Aria Digital Twin (ADT)

   Aria Digital Twin is an instrumented indoor (apartment + office) capture with ground-truth poses, depth and segmentation, released under the non-commercial Aria research license.

   Meta / Project Aria
6. Aria Everyday Activities (AEA)

   Aria Everyday Activities provides roughly 7.3 hours of everyday indoor activity with RGB, IMU, gaze and audio under the non-commercial Aria research license.

   Meta / Project Aria
7. EgoExoLearn: bridging egocentric and exocentric skill learning

   EgoExoLearn pairs egocentric and exocentric video of procedural task-following across daily-life and lab settings under a research license.

   OpenGVLab
8. Humanoid data: 10 Things That Matter in AI Right Now | MIT Technology Review

   MIT Technology Review documents the real-world-data bottleneck that makes consented first-person capture across environments urgent.

   MIT Technology Review

FAQ