imast.ai Now Records Real Work for Robots: Egocentric Video of Hotels, Kitchens and Short-Lets

Short version

• Imast AI records first-person video of real work in hotels, short-let apartments and kitchens.
• The footage is for teams that train robots and embodied AI: manipulation policies, world models and video-language models.
• Every worker signs a written consent. Every venue agrees. Faces and sensitive details are de-identified before storage.
• Video spec, metadata and delivery are set per buyer. You pay for the hours you accept, not the hours recorded.
• Start with a pilot of 20 to 50 hours of accepted footage.

Egocentric video is footage from the worker's own point of view, usually from a head-mounted camera. It shows hands, tools and objects the way the person doing the task sees them.

Robots are expected to work next in hotels, kitchens and apartments. To learn how hands get things done there, a model needs examples of real, unscripted work in those places. That is what Imast AI records. The work is not staged: people do their actual jobs in working venues.

Long tasks, done by hand, in the places robots are expected to work next:

• Hotel housekeeping. Bed making and linen change, bathroom and surface cleaning, amenity restocking, room reset checks.
• Kitchens. Prep, chopping and plating, dishwashing and racking, station clean-down, storage and restocking.
• Reception. Key and card handling, luggage, desk organisation, paper and device workflows.
• Laundry. Sorting and loading, folding sheets and towels, ironing, stacking and trolley packing.
• Short-let and Airbnb turnovers. A full apartment turnover between guests, kitchen and appliance reset, linen swap, and navigation in small rooms with different layouts.

Task lists can be agreed with you, so the coverage matches what your model must learn.

Robotics and embodied AI teams that need real physical work as training data:

• Teams that train robot manipulation policies.
• Teams that build world models.
• Teams that train video-language models.

The site also has a plain summary of what the Ego4D license allows, and a page on hotel robot training data.

Consent and privacy are steps in the process, not paperwork at the end:

1. Written worker consent. Every person who wears a camera signs an informed consent form, and can pause, stop or withdraw.
2. Venue agreement. The hotel, kitchen or property owner agrees to the recording, its scope, and the areas that stay off-limits.
3. De-identify before storage. Faces and sensitive details, such as documents, screens, room numbers and guest belongings, are blurred or removed before footage is stored.
4. Quality control. Each clip is checked against your spec, for framing, task coverage and continuity, before it is offered for acceptance.
5. Pay for accepted footage. You review deliveries. Billing is tied to the hours you accept.

Guests are not the subject. Imast AI avoids filming them, and any incidental people, faces or personal items are de-identified. Usage terms for each dataset are set out in the agreement with the buyer.

• Raw video. First-person footage of complete tasks, after de-identification. Resolution, frame rate and codec per your spec.
• Session metadata. Environment type, task description and session context, with fields agreed up front.
• Camera pose. Optional, where the capture setup supports it.
• Annotation. Task segments, actions or object labels, scoped to your taxonomy, on request.
• Delivery. File structure and transfer method agreed per project.

A pilot is a small, scoped batch of 20 to 50 hours of accepted footage. It lets you test the data in your own pipeline before you commit to volume.

1. Scope and spec. Environments, tasks, format and metadata fields.
2. Record and de-identify. Consented sessions, processed before storage.
3. QA and acceptance. You review samples and accept against the spec.
4. Decide on scale. Adjust the spec, or expand to more sites and tasks.

Pricing depends on environments, spec and volume, and is quoted per project.

Tell us what your model needs to learn. Request a sample or book a call on imast.ai, or email [email protected].