Egocentric Data Collection in Spain

Egocentric data is video recorded from the point of view of the person doing the work, with their hands, tools and the objects they handle in frame. Commercial egocentric data collection narrows that to paid, real-world work settings: the shelves, stockrooms, service bays, commercial kitchens and workbenches where the tasks an embodied AI or robotics model has to learn actually happen.

In Spain, Firsthand runs this as custom collection. You write down what the model is missing, which tasks, which environments, which conditions and how deep the annotation goes, and contributors in Spain are recruited to capture exactly that, in the commercial settings the spec names.

Geography is a diversity axis in its own right. Packaging, tools, store layouts and work habits differ from market to market, and a model trained on one country's footage quietly overfits to it. Adding Spain to a spec is how a program widens that spread on purpose instead of collapsing onto whoever was easiest to recruit.

Spain is one of the listed capture countries Firsthand publishes, but its consent framework is not yet documented in the provenance section, so no statute is printed and no environment is listed for it.

Commercial settings are recruited to spec, and the legal basis is confirmed per engagement and written into the spec you approve before capture starts.

"Listed" means the environment already appears for Spainin the published coverage data. "Recruited to spec" means contributors are sourced for that environment when a spec asks for it; pool sizes are quoted per engagement rather than published.

Household chores and home routines are excluded from this program. If your model needs home footage, see everyday routine collection instead.

The default tier for commercial collection in Spainis a recent-generation smartphone worn in a chest mount. It keeps both hands free and in frame, it is quick to deploy on a shop floor, and it records the task from roughly where the worker's eyes are. The project sets the minimum device spec, and contributors whose device does not meet it are not admitted to that project.

When a spec needs more than phone video, the same program can step up a tier: a 360 head mount for wider context, or the full instrumented rig with a head-mounted camera, dual wrist cameras, depth and IMU, synchronized to a shared clock. Same spec, same QA bar, whichever tier the project uses.

See egocentric video data collection for how the tiers compare, and sample data outputs for the files a recording ships with.

Spain is not yet one of the jurisdictions with a consent framework documented in Firsthand's provenance section, so no statute is printed here. That is deliberate: the site does not name a legal basis for a country it has not papered.

Instead, the legal basis and environment clearance for Spain are confirmed per engagement, before capture starts, and written into the spec you approve.

Releases are collected before capture, and contributors stop recording when an un-consented person enters the manipulation region. Every hour ships with its consent record, so the provenance of a clip can be traced back to a signed release.

Read the full chain of custody on the provenance page.

Commercial settings put bystanders, customers, vehicles and screens in frame far more often than a home does. Every clip runs through the anonymization pipeline before delivery: faces, licence plates and screens are detected, tracked across frames, redacted and then checked in a human review step.

Detection works on bounding boxes, so what gets blurred is auditable rather than a black box. The method is documented on the anonymization page.

Footage from Spain is delivered in the same formats as every other Firsthand program: RLDS, LeRobot, WebDataset, HDF5, zarr and Rerun .rrd, each mapped to the canonical episode schema. Annotation depth, from action segments to hand pose, is whatever the spec asked for.

Data ships under a buyer-owned commercial licence. See delivery formats for the schema and pricing for how a quote is built.