Two items, named as the catalog names them
"Check-in handling" reads "Handle documents and room keys." A spec names the document and the key as separate items and says what the hands do with each, rather than relying on one generic front desk label.
Now recruiting capture teams worldwide. Get paid to record first-person video of real work.
Apply to captureTask row
Why this row
"Check-in handling" reads "Handle documents and room keys." A spec names the document and the key as separate items and says what the hands do with each, rather than relying on one generic front desk label.
A key or a sheet is small and often covered by the fingers. 3D hand pose carries a per-joint visibility flag, so a finger hidden behind a document is marked as hidden rather than silently filled in.
The site says a staged working set recorded by a multiview rig, with the operator using handheld grippers, lets demonstrations map onto a robot end effector. That is the capture setup the row lists.
The row is Available on request, so it is scoped and recruited to spec when a buyer asks. It is not a claim that footage of this task has been delivered, and guest-facing settings carry a bystander and screen-blurring cost that the spec has to allow for.
Definition
It is video of a person handling documents and room keys, recorded to a written spec, with hand pose and action segments alongside. People reach it through several phrasings: egocentric front desk check-in data, first-person room key handling video, a handheld gripper document handling dataset, or custom hospitality data collection. They describe one request from different angles.
The site does not name a hotel, a front desk or a guest for this row. The catalog places it in a Hospitality Service Set described as a table, a counter or bar, and a housekeeping or room area, and the row wording is "Handle documents and room keys." This page uses "hotel check-in" as the search phrase for that wording and stays inside it. The data is human capture by paid, consented contributors, not robot or teleoperation data, so using it for a specific robot still needs your own retargeting step.
Catalog row
One. "Check-in handling" sits in the Hospitality Service Set (table, bar, room) environment of the Hospitality & Events sector. Its description reads "Handle documents and room keys." It is rated Easy, lists the capture setup "Multiview rig + handheld grippers", and has the status Available on request. The source is the site’s task catalog, in data/commercial-task-catalog.tsv and lib/commercial-catalog-data.ts.
Available on request means the work is scoped and recruited to spec when a buyer asks. It is not a statement that footage of this task has been delivered. The site states no delivered hours for the row, so this page claims none, and it names no document type, key type or desk layout because the catalog names none. The first table below reproduces the row exactly as the catalog states it, with neighbouring rows shown for context only.
The setting
The catalog describes the environment with three examples: a table, a counter or bar, and a housekeeping or room area. That is an environment type, not a named venue. The site does not say that any check-in capture takes place inside a particular hotel, reception or client facility, and this page does not claim it. The site also states that Firsthand does not capture at a customer’s own facility, so the set is a capture setting arranged for the task.
The site’s hospitality page does list a single reception or check-in interaction as an example of a bounded guest-facing task that fits a written spec. It also states that front-of-house hospitality settings are bystander-dense, which lowers yield after de-identification, and that hospitality is a thinner-coverage environment. A check-in spec has to allow for both, and the sections on privacy and speed below return to them.
Capture setup
The catalog lists "Multiview rig + handheld grippers". The site describes that mode as a staged working set recorded by a multiview rig while the operator performs the task with handheld grippers, so demonstrations map onto a robot end effector. The catalog names a second mode, a head-mounted phone or a multiview rig, for other environments, and the check-in row does not list it. This page therefore does not describe the row as head-mounted phone capture, and it does not describe it as stereo capture.
Egocentric is the name of this page family on the site, which treats first-person and operator-view capture as its default. How a head or operator view fits into the multiview rig for this row is not stated, and neither is a camera count, a resolution, a frame rate or a lighting setup. Those are agreed in the written spec rather than assumed. If a stereo pair matters to your policy, the site offers stereo only through its general stereo offer, agreed in the spec and confirmed on a Spec Pilot.
Handheld grippers
The catalog lists handheld grippers because the operator holds a gripper rather than a bare hand while performing the task, and the site says this lets demonstrations map onto a robot end effector. For documents and keys, that matters because picking up a thin sheet or a small key and passing it on are moves a policy has to reproduce with a gripper, not with five fingers. A handheld gripper document handling dataset keeps the demonstration closer to the hardware a robot would use than a bare-hand clip does.
The site does not state which gripper is used, how it attaches or how it holds a sheet or a key, and this page names none. It does not state a document size, a key size or a handover distance either. The site also does not claim that a model trained on this capture succeeds at the task, and it does not claim that human capture replaces robot data. The embodiment mapping is a design choice agreed in the skill spec.
Steps
The row names one verb, handle, and two items, documents and room keys. A spec that says "check a guest in" produces one long action segment. A spec that names the verb and the noun for each part produces a sequence you can slice by step. The site’s skill-spec guide says to name the verb and the noun rather than a broad activity, and the row’s own two items are the place to start.
Action segments are frame-accurate and share one clock with the video, so a boundary lines up with the frame where it happens. Contact events mark when fingers make and break contact with an item or a surface, and the spec decides what counts as the item: the document, the key, or the surface it rests on. Whether an episode covers one item, both items in one exchange or a run of several exchanges is a spec decision, because the site sets no episode length. This page gives no procedure for how check-in is done.
Difference
This page treats the document-and-key exchange as the unit of work. The egocentric mailroom and document handling page is organised around mail, filing, scanning, printing and document station rows in office environments, and its subject is paper work as a whole, not a room key or a guest-facing exchange. It is the right neighbour for what a handheld gripper does with paper, and it holds the site’s clearest statement on how documents are blurred.
The egocentric hotel room turnover page is organised around the Hotel & Lodging row "Room Turnover & Inspection", which is a back-of-house reset of a guest room and is rated Medium and Ready to deploy. The egocentric tray assembly and delivery page is organised around a different row in the same Hospitality Service Set, "Tray assembly + delivery". None of the three has check-in handling as its main topic, which is why this page exists. The second table below sets them side by side.
Neighbouring rows
The same environment lists Set table service ("Lay a full place setting."), Clear + bus table ("Stack dishes, bus, and wipe."), Pour + serve beverage ("Pour a drink and serve to a place."), Fold napkins ("Fold napkins to a set style."), Restock amenities ("Restock a minibar/amenity tray."), Tray assembly + delivery ("Arrange, carry, and set a tray.") and Wipe + reset counter ("Clear, wipe, and reset a counter."). These are separate rows with the same capture setup, and they are not steps of the check-in row.
A program that wants a longer hospitality routine can name several rows in one spec and say how episodes are divided. The catalog does not describe a combined routine, so this page does not claim that one exists. Each added row is its own scoping question, with its own status as the catalog lists it.
Streams and labels
The site states the episode schema for its custom collections: head and wrist RGB, depth, 3D hand pose, contact events and action segments, ten streams in all, aligned on one hardware clock and synced under 2 ms. Hand pose is 21 joints per hand with a visibility flag for each joint. The site’s only language data tied to Firsthand capture is one task instruction per episode.
The site does not state which of these streams the multiview rig with handheld grippers emits for the check-in row, so confirm the list in the spec and on a Spec Pilot. Instance masks are stated on other pages and not on the hospitality pages, so say so in the spec if your policy needs them. Object labels use the terms your spec sets, so they follow the vocabulary your training stack expects.
Privacy
Capture follows a documented consent chain for the operator wearing or holding the capture hardware, and the consent record is part of the delivery. Every episode links to a signed contributor release. Anyone else who appears in frame, such as a guest, a co-worker or a bystander, is handled through the site’s bystander and de-identification process rather than being captured as a second consenting participant. The data holds no guest names or guest identity documents.
The site states that faces, plates and screens are blurred irreversibly, that a reviewer confirms no un-consented person is identifiable, and that footage that cannot be de-identified is discarded. On its mailroom page the site describes the documents class as phone and monitor content, forms, badges and mail that would leak names, addresses or account data, which are irreversibly blurred and confirmed by a human reviewer. This page makes no wider claim for a check-in row. What may appear on a sheet, and who may appear in frame, is agreed in the spec before capture. This is not legal advice, and the executed agreement controls.
Not stated
Document type, key type and key mechanism are not stated. A hotel, a front desk, a guest, a room number and any property system are not stated. Outcome or goal-state labels, such as a completed check-in or a key handed over, are not stated, and neither are object identity labels, object 6-DoF pose, handover distance, time, success rate or throughput.
Force and tactile streams are not stated, so the data carries no reading of how firmly a key is held. If your policy needs any of these, they are defined in the written spec as annotation depth, and the quote covers them before capture begins. The third table below lists each item beside what the site says about it.
Hard cases
The site states that, once footage is captured, failure or recovery cases are captured and labeled where they occur. A dropped key or a document picked up again belongs in that category, because a policy that only sees clean takes has no model of recovery. The site gives no rate for how often these cases occur, and this page gives none.
Whether failures stay in a batch, and how a recovery is segmented, are spec decisions. QA also applies the site’s stated reject rules: an episode is rejected for a sync breach beyond the ceiling for more than three consecutive frames, for the manipulated object out of frame for more than 15 percent of the episode, for an undocumented bystander, or for a mismatch against the named skill spec. For this row, the spec decides whether the manipulated object is the document or the key. A reject log ships with every batch.
Scale and speed
The row is Available on request, so this is custom collection, not an off-the-shelf purchase. It is custom collection at scale through a network of thousands of capture providers, for a consistent, recurring data flow to labs. The site states a median of 48 hours from signed spec to first validated batch where coverage is deep, and three to four weeks where coverage is thin.
Hospitality is classed as a thinner-coverage environment, so a check-in order should expect the longer range, and the quote tells you which case you are in. On-demand top-ups run against the same spec and accept criteria after each training run, so new data stays consistent with what you already hold. A Spec Pilot, a small first batch, can prove the spec before you scale. Collections are billed per validated hour, and rejected episodes and recaptures are never billed.
Delivery and license
Delivery formats are RLDS, LeRobot and WebDataset, plus HDF5 and zarr. The first three are training-loader formats, and HDF5 and zarr carry arrays for custom pipelines. Readable converter source ships alongside the data so the schema can be retargeted to your own loader, and the episode schema page lists the fields.
The license is perpetual, irrevocable and worldwide, with the data buyer-owned and non-exclusive by default. Training, evaluating, fine-tuning and deploying models, including commercial ones, is permitted, and exclusivity is available as a separate term. The executed agreement controls. The site does not publish a price list, a minimum order or a batch size for this row, so those are set in your quote.
Ordering
Describe the documents and the keys you want in frame and where each exchange starts and ends. Say whether an episode covers one item, both items or a run of several exchanges, what counts as the object and what counts as a finished step, whether drops and recoveries stay in the batch, the privacy rule for the space and who may appear in frame, and the variation your pipeline is missing. Then name the streams you want, the delivery format your loader reads and whether a stereo pair is part of the request.
Most programs begin with a Spec Pilot, which proves the spec and accept criteria on real footage before you scale. The last table below shows what each spec element decides.
At a glance
Every collection runs through the same seven-stage end-to-end custom collection pipeline. End-to-end custom data collection is a managed service that takes an AI data need from problem to owned dataset in a single accountable pipeline.
Catalog row
| Task (as listed) | Description (as listed) | Environment | Skill | Capture setup | Status |
|---|---|---|---|---|---|
| Check-in handling | Handle documents and room keys. | Hospitality Service Set (table, bar, room) | Easy | Multiview rig + handheld grippers | Available on request |
| Tray assembly + delivery (context only) | Arrange, carry, and set a tray. | Hospitality Service Set (table, bar, room) | Easy | Multiview rig + handheld grippers | Available on request |
| Set table service (context only) | Lay a full place setting. | Hospitality Service Set (table, bar, room) | Medium | Multiview rig + handheld grippers | Available on request |
| Clear + bus table (context only) | Stack dishes, bus, and wipe. | Hospitality Service Set (table, bar, room) | Easy | Multiview rig + handheld grippers | Available on request |
| Restock amenities (context only) | Restock a minibar/amenity tray. | Hospitality Service Set (table, bar, room) | Easy | Multiview rig + handheld grippers | Available on request |
| Document type, key type, hotel, front desk, guest, outcome labels | None beyond the row wording | Not stated as rows | Not stated | Not stated | Not stated |
Compared pages
| Page | Rows it is organised around | Where check-in or room keys appear | Relation to this page |
|---|---|---|---|
| This page: egocentric hotel check-in and room key handling data | Check-in handling (Easy, Available on request) | Main topic | Documents and room keys as the unit of work |
| Egocentric mailroom and document handling data | Mail handling, filing, scanning, printing and document station rows | Not a topic; paper work in office environments | Different main task; the neighbour for paper handling and document blurring |
| Egocentric hotel room turnover data | Room Turnover & Inspection (Medium, Ready to deploy) and neighbouring hotel rows | Not a topic; back-of-house reset of a guest room | Different main task; room reset, not a guest-facing exchange |
| Egocentric tray assembly and delivery data | Tray assembly + delivery (Easy, Available on request) | Not a topic; same environment and capture setup | Different row; arranging, carrying and setting a tray |
| Hospitality egocentric data | Hospitality and housekeeping work as a whole setting | Lists a single reception or check-in interaction as a guest-facing example | The broader setting page |
Capture setup
| Item | What the site states | For a check-in order |
|---|---|---|
| Multiview rig + handheld grippers | A staged working set is recorded by a multiview rig while the operator performs the task with handheld grippers, so demonstrations map onto a robot end effector | The mode the check-in row lists |
| Phone (head-mounted) / multiview rig | A head-mounted phone records the first-person view, and a multiview rig adds synchronized third-person angles when the spec asks for them | Not listed for the check-in row; named in the catalog for other environments |
| Stereo capture | Offered through the general stereo offer, agreed in the skill spec | No stereo pair is stated for the check-in row; a spec choice |
| Episode schema | Head and wrist RGB, depth, 3D hand pose, contact events and action segments; ten streams synced under 2 ms | Stated for custom collections; which streams the row emits is Not stated |
| Hand pose | 21 joints per hand with a per-joint visibility flag | Stated |
| Camera count, resolution, frame rate, lighting | Not stated for hospitality rows | Agreed in the spec |
| Gripper model, attachment, how a sheet or key is held | Not stated | Not claimed; agreed in the spec |
| Failure and recovery cases | Captured and labeled where they occur | Stated; how often they occur is Not stated |
Stated vs not stated
| Item | What the site states | For a check-in order |
|---|---|---|
| Row wording | "Handle documents and room keys." | Stated; documents and room keys are the two items |
| Skill and status | Easy; Available on request | No volume, hours or delivered footage is claimed on this page |
| Environment | Hospitality Service Set: a table, a counter/bar, and a housekeeping/room area | An environment type, not a named venue |
| Hotel, front desk or guest setting | Not stated for this row | Not claimed; set in the spec |
| Document type, key type, key mechanism | Not stated | Not claimed; set in the spec |
| Outcome, goal-state or completion labels | Not stated | Not claimed; set in the spec if needed |
| Object identity labels, object 6-DoF pose | Not stated | Not claimed; set in the spec if needed |
| Force or tactile data | Not stated | No grip reading is described as delivered |
| Instance masks | Stated on other pages, not on the hospitality pages | Confirm in the spec if needed |
| Task duration, success rate, throughput | Not stated | Not claimed on this page |
| Hospitality coverage | Thinner; a first batch typically takes three to four weeks | Expect the longer range; the quote confirms |
| Faces and screens | Blurred irreversibly; a reviewer confirms; footage that cannot be de-identified is discarded | Stated; who may appear in frame is agreed in the spec |
| Documents | On the mailroom page: forms, badges and mail that would leak names, addresses or account data are irreversibly blurred and reviewed | Stated there; no wider claim is made for this row |
| Bystanders in front-of-house settings | Bystander-dense, which lowers yield after de-identification | Stated |
Formats
| Format | Used for | Notes |
|---|---|---|
| RLDS | Training loader; TFDS-compatible tfrecord shards | Episode and step structure |
| LeRobot | Training loader; parquet plus mp4 | Named in the site’s format list |
| WebDataset | Training loader; tar shards for streaming | Named in the site’s format list |
| HDF5 and zarr | Arrays for custom pipelines; one group per episode | For custom loaders |
| Converter source | Readable source ships with the data | Lets you retarget the schema to your own loader |
Skill spec
| Spec element | What it decides | Where it shows up |
|---|---|---|
| Row | Check-in handling, worded as the catalog words it | Set per program; Available on request as listed |
| Items and steps | What one step is for the documents and for the room keys | Episode boundaries and action segments; set per program |
| Episode scope | One item, both items or a run of several exchanges | Set per program; the site does not state an episode length |
| Object definition | Whether the document or the key is the manipulated object | Contact events and the out-of-frame reject rule; set per program |
| Setting | Which of the table, counter/bar or room area is in frame | Set per program; the site names no venue |
| Rig and streams | Multiview angles, wrist views, hand pose, and any stereo pair | The channels you can read; confirm on the Spec Pilot |
| Failures | Whether drops and recoveries stay in the batch | Labeled where they occur; set per program |
| Delivery format | How the data is packaged | RLDS, LeRobot, WebDataset, HDF5 or zarr |
| Accept criteria | What passes QA, including what counts as a reject | Reject log per batch |
| Privacy | Who may appear in frame and what a sheet may show | Faces and screens blurred at ingest; other items agreed in the spec |
| License | Who owns the dataset | Perpetual, irrevocable, buyer-owned, non-exclusive by default |
Example footage
One real clip already published on this site, shown as a format example: a first-person view with faces blurred. It does not show check-in, documents, room keys, a hospitality set or the multiview rig with handheld grippers that the check-in row lists.
No check-in or room key clip is published on the site, so this is labelled as a format example. Request a sample to see delivered files. See sample data outputs, request an egocentric data sample.
FAQ
Sources
Tell us the documents and keys you want in frame, whether an episode covers one item, both items or a run of exchanges, what counts as the object, which part of the service set is in frame, whether drops and recoveries stay in, the privacy rule for the space and the format your loader reads. We will quote reach, timeline and price against a written spec.