Tidy-Up Kitchens Explorer

Dataset in action · Real → Sim

Captured by people.
Rebuilt for robots.

Harvard’s Computational Robotics Group used our kitchen captures to build a robot simulation environment with SceneAgent, turning a real home into a scene with articulated cabinets and appliances.

The reconstructed CollectData kitchen with a Dexmate robot — A CollectData kitchen reconstructed by SceneAgent.

SceneAgent · Harvard Computational Robotics Group
Luke Hollis, Tianxing Fan & Heng Yang.
Kitchen captures by CollectData. Reconstruction and simulation by SceneAgent.

Explore the SceneAgent project ↗

Provenance & rights

Cleared for model training.

Payout ledgerone kitchen

Before — messy
1	Scan your messy kitchen	PAID
2	Widest possible photo of your messy kitchen	PAID
3	Photo of the messiest counter	PAID
4	Label the messy counter	PAID
5	Closeup photo #1	PAID
6	Label photo #1	PAID
7	Closeup photo #2	PAID
8	Label photo #2	PAID
9	Closeup photo #3	PAID
10	Label photo #3	PAID
11	Messy kitchen video tour (naming items to tidy)	PAID
After — tidied
12	Photo of the clean counter	PAID
13	Scan your clean kitchen	PAID
14	Widest possible photo of your clean kitchen	PAID
	14 tasks, one kitchen	PAID

LiDAR, photo, labeling and video tasks — each priced, reviewed and paid on its own. Nothing ships until its contributor is paid.

Commissioned, not scraped Every capture was shot to our spec by a contributor we recruited and paid. One clean chain of title, nothing licensed in from a third party.
Signed release on every capture A perpetual, worldwide, sublicensable right to use the media for training, fine-tuning and evaluating AI models, and to create derivative works.
US contributors, US homes US-based adults capturing their own homes, identity- and payment-verified through our payout provider.
Privacy cleared before delivery Faces blurred across the set; captures we could not clear are withheld rather than shipped.
Commercial training licence Non-exclusive: no attribution, no copyleft, no per-model royalty, no limit on deploying what you train. Exclusivity available on request.
Auditable per capture Release version, acceptance timestamp and redaction decisions ship as provenance metadata alongside the data.

Need the paperwork before the data? Ask us for the release text and licence terms →

Beyond kitchens

We capture other domains too.

Fridge interiors

LiDAR scans and guided photo passes of open refrigerators - shelf-level geometry with per-item photos, for inventory and grasping datasets.

Egocentric laundry

Head-mounted first-person video of full laundry routines - sorting, loading, folding - with task narration, for household-robotics training.

Kitchen #01 · everything we captured

Talk to us Download the dataset

MESSY

CLEAN

photogrammetry model · drag to orbit

No kitchen video captured for this kitchen yet.

LiDAR mesh · drag to orbit

reconstructed room + placed objects · drag to orbit

All captures from this kitchen Thumbnail size

All kitchens Thumbnail size

Jump to another kitchen →

3D scans of real kitchens - messy and clean.

Captures from real kitchens. Pick one to explore.

Captured by people.
Rebuilt for robots.

More than a scan.

Cleared for model training.

We capture other domains too.

Fridge interiors

Egocentric laundry

How we captured the kitchens.

Talk to us

Publish policy

Unit selection

Asset classes

Redactions

3D scans of real kitchens - messy and clean.

Captures from real kitchens. Pick one to explore.

Captured by people.Rebuilt for robots.

More than a scan.

Cleared for model training.

We capture other domains too.

Fridge interiors

Egocentric laundry

How we captured the kitchens.

Publish policy

Unit selection

Asset classes

Redactions

Captured by people.
Rebuilt for robots.