Skip to content

Egocentric video data acquisition, executed end to end.

For physical-AI and robotics programmes, we deploy trained recorders on site with wearable capture rigs, our own or yours. They capture day-to-day work or perform the tasks you specify, supported by in-house IT, annotation, narration and quality-control teams. Engagements are invoiced solely on approved hours.

Hands shaping clay on a potter’s wheel (illustrative stand-in image, not SankhyaData footage).

Narrator: Pressing in with both thumbs to narrow the neck.

Illustrative stand-in image, not SankhyaData footage; annotation, narration and recorder HUD shown schematically.

On-site capture across some seventy operational environments.

Schematic: each point stands for ten hours of the 50,000 handled.

50,000+hours

of egocentric footage handled by our teams, recorded on site during day-to-day operations.

2,000approved hours a week

deliverable through our field contributor network.

Figures from our own records.

Examples from some seventy operational environments, by sector

Repair and servicing
  • Appliance repair
  • Bicycle repair
  • Car detailing
  • Car repair
Food production
  • Bread bakery
  • Cake shop
Manufacturing
  • Building materials manufacturing
  • Metal fabrication
Agriculture
  • Agricultural farm

Every hour is a path through a real place.

A recorder moves through the capture site, and the camera records what the recorder sees. Scroll, and the view passes from the site into the recorder’s own eyes.

Schematic: a recorder’s head trajectory and camera field of view in a workshop, drawn as a point cloud.

One accountable team, from scoping to delivery.

  1. 01

    Scoping

    We translate your task taxonomy, target operational environments and required hours into a scoped capture plan.

  2. 02

    Instrumentation

    Capture rigs, our own or client-supplied, are provisioned by our in-house IT team before deployment, then fitted to each recorder and task.

  3. 03

    Acquisition

    Trained recorders capture first-person footage of real work in situ, or perform the tasks the client scripts, across the operational environments the client specifies.

  4. 04

    Annotation

    Our annotators author scene descriptions and temporally aligned subtitles in strict conformance with the client’s guidelines.

  5. 05

    Narration

    Our narrators record the scene descriptions and subtitles wearing active noise-cancelling headphones, after a microphone and ambient-noise check.

  6. 06

    Quality control

    Each submission receives an AI-generated quality score; our quality-control (QC) reviewers adjudicate low-scoring items and audit one approval in ten.

  7. 07

    Delivery

    Approved hours are released in numbered batches through the authenticated client portal, each with a checksum manifest and a verification script.

An in-house IT function underpins every engagement.

Our own IT engineers provision every device, prepare each capture site and stand by to assist field officers and recorders the moment an issue arises.

IT hubCapture siteIT at workSchematic: sites and timings are illustrative.
Device provisioning
Every rig is configured, updated and charged by our IT staff before deployment.
Site readiness
Each capture site is prepared and checked by our IT staff before recording begins.
Client portal
Authenticated access to review queues and delivery batches.
Immediate field support
Our IT team is on call to field officers throughout every capture session, resolving device and recording issues at once, on site or remotely.

Four capture rigs, each fitted to the recorder and the task.

Schematic: where each rig sits and what its camera sees. Scroll through the rigs, or select one, to fit it.
Rig
Head-mounted iPhone
Mount
Head mount
Vantage
Eye level
Notes
Records along the recorder’s line of sight.
Rig
Cap camera, power bank
Mount
Camera on the cap, powered by a power bank
Vantage
Brow level
Notes
Records from just above the recorder’s eyeline.
Rig
Cap camera, phone in pocket
Mount
Camera on the cap, USB-C to the phone
Vantage
Brow level
Notes
The cap camera connects by USB-C to a smartphone in the recorder’s pocket, which records through the capture app.
Rig
Human rig
Mount
Wearable rig
Vantage
Per task
Notes
Our preferred configuration, deployed wherever the task permits.

Our rigs record 1080p at 30 fps, encoded in HEVC (H.265), and are provisioned by our in-house IT team.

Two further arrangements, available on request

  • Client-supplied devices

    Where a programme specifies its own capture hardware, such as smart glasses or a proprietary head rig, we take custody of the client’s devices. Our IT team provisions and maintains them for the duration of the engagement, and our recorders capture with them under the same methodology and quality control. Our teams have worked with established capture and recording platforms.

  • Directed tasks

    Beyond footage of day-to-day work, our recorders perform the specific tasks a client scripts, in the environments it specifies, so that every approved hour falls within the task taxonomy.

What each delivery batch contains.

Batches are numbered and released through the authenticated client portal.

  1. Video

    First-person footage, 1080p at 30 fps, HEVC (H.265)

  2. Scene descriptions

    Authored by our annotators in strict conformance with the client’s guidelines

  3. Subtitles

    Temporally aligned to the footage

  4. Voice-over

    Recorded by our narrators for every scene

  5. Organisation

    By operational environment, capture site and date

  6. Manifest

    Duration, size and MD5 checksum for every file

  7. Verification

    A verification script supplied with every batch

  8. Released as a numbered batch through the client portal

Invoicing is contingent on approved hours.

Each recorded hour is reviewed against the agreed specification. Only approved hours are counted, invoiced and delivered.

  1. RecordedCaptured on site
  2. ReviewedAI quality score, then QC adjudication
  3. ApprovedCounted and invoiced
  4. DeliveredIn numbered batches with a checksum manifest

Footage is held in confidence and delivered with verifiable integrity.

Confidential handling
Client footage is handled as confidential material; clients access it only through authenticated portal accounts.
Verifiable delivery
Every batch ships with a manifest recording each file’s duration, size and MD5 checksum, together with a verification script.
Durable custody
Stored footage is protected by verified nightly backups.
Audit trail
The handling of client footage is documented in tamper-evident audit records.

About one representative hour from each operational environment.

Review our capture standard before commissioning: raw first-person recordings with their metadata, viewable in the browser through our samples viewer.

Schematic: the representative set, one tile per operational environment (69 in all). Named tiles are examples; interiors are abstract, not footage.

Then send us your brief

Questions asked before commissioning.

What does a capture programme include?

Egocentric video data collection from scoping to delivery: wearable capture rigs (our own or client-supplied), trained recorders performing day-to-day work or directed tasks, and our in-house annotation, narration, quality-control and IT teams.

Which capture rigs do you deploy?

Four rigs: a head-mounted iPhone; a cap camera powered by a power bank; a cap camera connected by USB-C to a phone in the pocket, which records through the capture app; and the human rig, our preferred configuration. All record 1080p at 30 fps, encoded in HEVC (H.265).

Can you capture with our own devices?

Yes. We take custody of client-supplied devices; our IT team provisions and maintains them, and our recorders capture with them under the same methodology and quality control as our own rigs.

Can you work with our recording platform?

Yes. Our teams have experience with established capture and recording platforms, and our IT team prepares devices and recorders for the platform a programme specifies.

What happens if a device fails in the field?

Our IT team is on call to field officers throughout every capture session and resolves device and recording issues at once, on site or remotely.

Can your recorders perform tasks we specify?

Yes. In addition to recording day-to-day work, our recorders perform the tasks you script, in the operational environments you specify.

What volume can you deliver?

Through our field contributor network, we can deliver 2,000 approved hours a week. Only hours that pass quality control are counted, invoiced and delivered.

In which environments is footage recorded?

On site, during day-to-day operations, across some seventy operational environments, from agricultural farms to metal fabrication.

Which layers do you produce?

Scene descriptions and temporally aligned subtitles, authored by our annotators to your guidelines, and a voice-over recorded by our narrators for every scene.

How is an hour approved?

Every submission receives an AI-generated quality score. Our QC reviewers adjudicate low-scoring items and audit one approval in ten. Only approved hours are counted, invoiced and delivered.

How can we verify a delivery?

Each numbered batch carries a manifest recording every file’s duration, size and MD5 checksum, together with a verification script that checks each file against it.

Can we review footage before commissioning?

Yes. Our representative set holds approximately one hour from each of 69 operational environments, viewable in the browser through our samples viewer.

How does an engagement begin?

Write to contact@sankhyadata.com with your task taxonomy, target operational environments and required hours. We respond with a scoped capture plan and representative samples. To talk it through first, book a 30-minute call at calendly.com/sankhyadata.

Commission a capture programme.

Send us your task taxonomy, target operational environments and required hours. We will respond with a scoped capture plan and representative samples.