Skip to main content

Dexset

Real-world training data for physical AI

The real-world data engine for robots that work.

We collect, narrate, annotate and deliver the multimodal datasets that teach robots to act. Captured across real industrial, clinical, domestic and field environments, and quality-checked before it ever reaches your model.

Ego array · industrial assembly · multi-view

Capturing

At a glance

Proven at the scale physical AI demands.

A quality-checked library you can license today, plus live capture crews that stand up a bespoke program and scale it to a thousand hours a day.

Hours of OTS data

0 +

Off-the-shelf and quality-checked. Ready to license now.

Hours narrated

0 +

Industrial work, captured with synchronized audio and transcripts.

Hours per day

0 +

Live collection capacity across our active regions.

Countries

0 +

India, Latin America, Europe and Africa.

Off the shelf

License from a cleared, documented library and start training immediately.

Built to order

Commission capture for your exact tasks, environments and edge cases.

Multimodal

Mono and stereo video, narrated audio, depth, IMU and dense labels.

Compliant

Consent-based sourcing aligned to the law of every region we work in.

Why now

Physical AI is leaving the lab, at scale.

Humanoids are walking onto factory floors. Autonomous systems are moving through warehouses, roads and hospitals. These systems do not learn from clean theory; they learn from messy reality. The bottleneck has shifted from algorithms and hardware to the real-world data that teaches a robot to act.

Pilots are becoming production

Robotics programs are crossing from demos into deployments where reliability is non-negotiable.

Models are going embodied

Vision-language-action and world models need diverse, action-grounded demonstration data.

Demand outpaces supply

Teams can capture far fewer real hours than their models actually need to generalize.

The problem

Raw footage is not training data.

Why robotics data is hard

  • It is multimodal. Video, depth, audio, IMU and joint states all have to line up on one clock.
  • It is action-grounded. Labels must match, to the frame, what the operator actually did.
  • It is expensive to capture. Real rigs and real environments cap how many hours a team can log.
  • The long tail matters most. The rare recoveries and failures are the hardest clips to collect.

What it costs teams

Models fail on contact with reality

Without edge-case and failure coverage, a policy that looks great in the lab breaks in the field.

Engineers become data plumbers

Time that should go to the model goes to capture, cleaning and labeling instead.

Bad labels fail silently

Misaligned data quietly degrades a policy in ways that are painful to debug later.

What we do

We turn real-world activity into training-ready data.

One partner for the whole job: define the task, capture it in the real world, narrate and annotate every action, check the quality, and deliver it straight into your pipeline.

End-to-end partner

0

Signal types captured

0 +

Real environments

0 +

Stage QA pipeline

0

The pipeline

One pipeline, from raw reality to your stack.

Step 1

Collect

Multimodal capture across real environments.

Step 2

Narrate

Operators describe each action, on the clock.

Step 3

Annotate

Dense, action-grounded labels and language.

Step 4

Validate

Five-stage QA on every batch.

Step 5

Deliver

Pushed straight into your training stack.

Input

Raw, real-world activity in.

Output

Training-ready datasets out.

Traceable

Every stage is versioned and auditable, so data is traceable from capture to delivery.

Capabilities

One partner from capture to training-ready data.

Capabilities · 1 of 4

Data collection

Purpose-built capture programs, designed around the behavior you want your model to learn. We agree on the tasks, environments and edge cases up front, then capture them at scale across synchronized cameras and sensors.

Synchronized streams

Frame-accurate alignment across every sensor.

Scenario-targeted

Built around your capability and benchmark goals.

Diversity by design

Environments and operators that generalize.

Rich metadata

Timestamps, activity labels and descriptors.

Capabilities · 2 of 4

Annotation

Dense, robotics-grade labeling that goes well beyond boxes on images. We annotate the sensor streams, actions and language that vision-language-action models actually consume, and a human expert reviews every batch.

Action-grounded

Aligned in time to what the operator did.

Multimodal grounding

Vision, language and action linked together.

Consistent taxonomy

Stable label schemas across every batch.

Human-in-the-loop QA

Expert reviewers keep accuracy high.

Capabilities · 3 of 4

Teleoperation

The gold standard for demonstration data. Skilled operators drive robots through target tasks and produce the exact state-action pairs that imitation learning consumes, with no inverse dynamics and no reward engineering.

State-action pairs

Directly consumable by imitation learning.

Throughput at quality

Scaled without sacrificing signal.

Benchmark-aligned

Mapped to your evaluation up front.

Iterative delivery

Ongoing drops to evaluate policies.

POST /v1/datasets/export  200 OK

{
  "dataset": "industrial_assembly_v3",
  "format":  "lerobot | rlds | coco",
  "signals": ["rgb","stereo","audio","imu"],
  "delivery": "stream",
  "qc_score": 96.4
}

Capabilities · 4 of 4

Delivery & formats

Data should not arrive as a zip file over email. Our delivery APIs push training-ready data straight into your environment, with the structure and metadata your tooling already expects.

10+

Export formats

3

Cloud targets

<1s

Stream latency

What we capture

The signals your model learns from.

Every clip we deliver is more than video. It is a stack of synchronized signals, aligned to the frame, so perception and vision-language-action models can consume it directly.

Mono RGB video

Single-camera egocentric view at 1080p, 30 fps.

1080p · 30 fps · egocentric

Stereo & depth

Synced stereo pairs with metric disparity and depth.

65 mm baseline · metric depth

Narrated audio

Operators describe each action, with clean transcripts.

Speech + timed transcript

Ego-motion / IMU

6-axis IMU at 300 Hz, synced to the camera to 1 µs.

6-axis · 300 Hz · 1 ms sync

Hand & grasp

Hand keypoints, grasp points and contact events.

Keypoints · grasp · contact

Action segments

Tasks split at the moment the operator’s intent changes.

Intent-boundary segments

Object grounding

Objects detected, labelled and located in frame.

Detect · label · locate

Scene & 3D

Scene geometry, multi-view and environment tags.

Multi-view · scene tags

Capture hardware

Mono and stereo rigs, built for the field.

Lightweight, cap-worn capture units that an operator can wear through a full shift. Choose a single-camera mono rig for reach and volume, or a calibrated stereo rig when your model needs metric depth.

Specification

Stereo rig

Mono rig

Camera configuration

2× RGB stereo

1× RGB

Per-camera resolution

1920×1080 @ 30 fps

1920×1080 @ 30 fps

Native FOV

120° H / 80° V / 150° D

120° H / 80° V / 150° D

Post-rectification FOV

118.17° / 61.47° / 143.76°

—

Baseline

65 mm

n/a · single camera

Shutter

Global

Global

IMU

6-axis @ 300 Hz

6-axis @ 300 Hz

Camera-IMU sync

1 µs

1 µs

Recording

On-device / SD card

On-device / SD card

Power

USB power bank

USB power bank

Weight

127 g

~95 g

Form factor

Integrated cap-worn

Integrated cap-worn

Stereo values are production spec. Mono rig figures marked “—” or “~” are indicative; final mono specifications available on request.

1 µs

Cam-IMU sync

300 Hz

IMU rate

127 g

Stereo weight

Industrial expertise

Ten thousand hours of narrated industrial work.

Operators narrate each task aloud as they perform it, so every clip ships with synchronized audio and a clean transcript. Your model learns the intent behind the motion, not just the motion.

10,000+

Narrated hours

Audio

+ transcript per clip

60 fps

High-speed lines

Factory · pipe-fitting · narrated

Experience

Every environment a robot has to work in.

High-speed packaging

Industrial · 60 fps

Precision machining

Skilled trades · lathe

Volumetric imaging

Clinical · 3D CT

Surgical endoscopy

Clinical · in-vivo

Humanoid locomotion

Embodied · Unitree G1

Kitchen manipulation

Domestic · hand pose

Mobile teleoperation

Field · outdoor

Structural handling

Construction · heavy

Sourcing

How we source data, responsibly.

Five ways to get the data you need, from a library that is ready today to a crew that captures the exact scenario your model keeps failing on. Every route is consent-based and documented.

Managed field crews

Trained operators capture on location in each region, scaling to 1,000+ hours a day.

Consented contributors

Every participant opts in. Consent, rights and usage are recorded per clip.

Partner sites

Factories, workshops and clinics captured under formal agreements.

Controlled studios

Repeatable rigs for targeted tasks, edge cases and benchmarks.

Off-the-shelf library

250,000+ hours already collected, cleared and ready to license.

Edge-case commissions

We go and capture the rare scenarios that your current data misses.

Speed

From brief to data, fast.

Because our crews, rigs and library are already in place, a new program moves quickly. You see a validated pilot batch in about two weeks, then we ramp capacity from there.

01

Scope & consent

We align on tasks, environments, edge cases and data rights.

Day 0

02

Rig & crew

Devices configured and operators briefed in the target region.

Week 1

03

First frames

A validated pilot batch lands in your pipeline for review.

~Week 2

04

Scale up

Capacity ramps toward a thousand hours a day.

Ongoing

05

Iterate

Fresh drops as your model and benchmarks evolve.

Continuous

Quality

Quality is a process, not a promise.

Bad data fails silently, so we build verification into every stage. Issues surface before data reaches your model, and every batch stays auditable from source to delivery.

Defined QA gates

Acceptance criteria agreed up front and checked on every batch.

Temporal & sensor checks

Alignment verified across modalities and action boundaries.

Expert human review

Skilled reviewers verify labels a model alone would miss.

Versioning & traceability

Full provenance from capture to delivery for audit and rollback.

Trust & compliance

Privacy and consent, built in.

Faces, plates and other identifiers are automatically redacted before data leaves our pipeline.

Consent-first

Every contributor opts in; rights recorded per clip.

PII redaction

Automatic face and identifier blurring on capture.

Encrypted handling

Secure transfer and storage, access-controlled.

Data residency

Regional handling aligned to local law.

Full provenance

Every dataset versioned and auditable.

Right to withdraw

Contributors can revoke; data is traced and removed.

Global reach

0 +

countries, four active regions.

A distributed footprint gives your models the environmental and cultural diversity they need to generalize, each region captured under its own data-protection regime.

India

DPDP Act 2023

Latin America

LGPD · regional law

Europe

GDPR · UK GDPR

Africa

POPIA · NDPR

Partners

Who we build data for.

We work with the teams building physical AI, from early-stage startups to the labs setting the frontier. Client names are withheld under NDA; references are available on request.

YC-backed startups

Early teams that need production-grade data without building a collection org first.

Frontier AI labs

Groups training large multimodal and world models on real-world signals.

Robotics firms · US & Europe

Manufacturers and integrators shipping manipulation and mobile platforms.

Humanoid & VLA teams

Groups that need teleoperation and action-grounded demonstration data.

Surgical & medical robotics

Clinical capture handled with the privacy and rigor the domain requires.

Autonomous systems

Perception and navigation teams across field, logistics and AV programs.

Differentiation

Why teams choose dexset.

01

Real environments, not stock

Industrial, clinical, domestic and field footage from the places robots actually work.

02

Narrated, action-grounded

Audio and transcripts carry the intent behind every motion, aligned to the frame.

03

Our own capture hardware

Mono and stereo rigs, sensor-synced to the microsecond, built for long field shifts.

04

Off-the-shelf and bespoke

License 250,000+ ready hours today, or commission exactly what you are missing.

05

Quality you can audit

Defined QA gates, expert review and full provenance on every batch.

06

Consent & compliance built in

Sourced and handled to the law of every region we operate in.

FAQ

Frequently asked questions.

What does dexset actually deliver?

dexset delivers ML-ready robotics training datasets: real-world video captured to your task brief, annotated with objects, actions, poses, task states, and failures, validated for coverage and consistency, and exported in COCO, YOLO, Pascal VOC, JSON, CSV, or a custom schema mapped to your pipeline.

Generic vendors label footage you already have. dexset runs the full loop: task definition, real-world capture with trained operators, robotics-specific annotation, quality scoring, and delivery. Labels carry physical context — grasp points, task states, failure markers — that general-purpose labeling teams routinely miss.

No — we complement it. Simulation is excellent for scale and rare-state coverage, but deployment requires real-world variation: lighting, clutter, occlusion, and human behavior. Most teams use dexset data to fine-tune, validate sim-to-real transfer, and build held-out evaluation sets.

A pilot of under ten hours of captured data typically ships in two to three weeks, including annotation and quality review. Larger programs run as recurring batches, so your first training-ready data arrives early rather than at the end of the engagement.

You do. Custom-captured datasets are delivered under exclusive ownership or exclusive license terms agreed before capture begins. Every clip carries source, consent, and capture records, so provenance is auditable from raw footage through to the exported training file.

Let's work together

Let's build the data behind your robots.

Tell us what you’re training and we’ll bring a data plan. A dexset data expert replies within one business day.