Frontier modalities
Egocentric data providers: an evidence-based comparison
Compare catalog claims, custom sourcing, closed-loop programs and technical evidence rather than ranking unverified inventory.
- Topic
- Frontier modalities
- For
- Buyers and data owners
- Reading time
- 5 min
- Updated
This comparison is published by Fiund, a commercial data-licensing business. It is not independent purchasing research, an inventory audit or an endorsement. We compare dated official descriptions and identify questions still requiring a sample or quote.
Provider, research dataset and contributor program are different categories. A recording program feeding one company’s system is not evidence that its data can be purchased on an open marketplace.
| Question | Evidence or decision |
|---|---|
| DataClue | Vendor catalog and project quote; verify the actual package |
| Troveo | Library and bespoke sourcing claims; confirm exact subset |
| Figure Index | Contributor program; no open external dataset sale established |
| Fiund | Brief and assessment routes; availability confirmed after review |
DataClue: a public catalog and project quotes
On the checked date, DataClue published scenario listings with prices and described stereo or multi-camera material as quoted per project. These are the vendor’s statements and asking prices; Fiund has not verified the delivered hours, technical evidence or completed transactions.
Request a representative manifest, sample, scope-specific quote and calibration or timing records for the claimed product. Confirm unique activity versus stream-hours and the source and review status of labels before using a catalog total in a buying decision.
Troveo: library and sourced-to-brief descriptions
Troveo’s site describes existing libraries and bespoke sourcing. That positioning is relevant when comparing already-recorded material with a requested collection. Its broad library claims do not establish availability for your exact tasks or geometry requirements.
Ask whether the proposal is an existing reviewed subset, material still needing clearance or a new collection. Require the actual formats, inclusion rules, preparation scope and pilot evidence rather than treating aggregate marketing totals as an accepted delivery.
Figure Index and research projects serve different roles
Figure Index describes a contributor program for real-world task capture. That page is useful for understanding a collection channel, but does not establish an open dataset sale to external buyers. Do not list it as purchasable inventory without an explicit offer.
UMI and HOT3D are research references for demonstration methods and hand/object tracking. Inspect their terms and formats directly. Public research access does not automatically permit commercial resale or make Fiund an authorized distributor.
Where Fiund fits and what remains to verify
Fiund accepts buyer briefs, existing-data assessments and collection-capability descriptions. Availability and the proposed deliverable need confirmation after review. This page makes no claim that Fiund has more hours, lower prices or better model performance than these organizations.
Use the same evidence request for every candidate: task match, source authority, original files, timestamp and geometry documentation where required, annotation provenance, acceptance policy and permitted use. Pick a provider after a relevant pilot, not from this page’s ordering.
Sources and further reading
- DataClue official catalog — checked October 27, 2026dataclue.app
- Troveo official site — checked October 27, 2026troveo.ai
- Figure Index — checked October 27, 2026figure.ai
- UMI research project — checked October 27, 2026umi-gripper.github.io
- HOT3D project — checked October 27, 2026facebookresearch.github.io
Frequently asked questions
Is this a ranking of the largest providers?
No. It is a source-based comparison of offer types and unresolved diligence questions, with Fiund’s commercial affiliation disclosed.
Are the advertised datasets independently audited here?
No. Official marketing descriptions are attributed as claims, not verified delivery evidence.
Have a specific collection requirement?
Send your requirements →Related resources
7 min read
How to evaluate a training-data vendor
A repeatable way to judge a vendor on provenance and rights auditability, not just samples and turnaround.
5 min read
Off-the-shelf versus custom egocentric collection
Choose an existing subset, new recording or a hybrid only after reviewing task coverage, missing signals and pilot cost.
5 min read
Egocentric video versus teleoperation versus UMI data
Separate observations of human activity from robot actions and gripper-based demonstrations before choosing training data.
5 min read
Egocentric data collection pay by program and country
Check dated official program terms, eligibility and billing units without turning unrelated job rates into footage prices.
5 min read
Recording workplaces for AI: permissions and preparation
Review site access, workers, bystanders, confidential material and intended use before recording a workplace for a dataset.
5 min read
Narrated egocentric data: collect useful task descriptions
Keep spoken narration, transcripts, action segments and generated descriptions distinct and linked to the original timeline.