> For the complete documentation index, see [llms.txt](https://docs.roboset.org/llms.txt). Markdown versions of documentation pages are available by appending `.md` to page URLs; this page is available as [Markdown](https://docs.roboset.org/atlas/overview.md). # Overview ### Overview Atlas is Roboset's data marketplace — the bridge between the contributor network and Physical AI buyers. It handles contribution, verification, enrichment, packaging, and licensing of real-world task data. It is live with over 1,000 hours of demonstration data, more than 500 active contributors, and 20 task categories spanning egocentric daily activity, cooking, manufacturing, warehouse, driving, and general human activity capture. ### The Four-Stage Workflow Every contribution moves through the same pipeline. Each stage adds the value that separates a home video from a licensable asset. 1\. Capture. Contributors film natural tasks from a first-person perspective — chest-mounted phone, head-worn camera, or webcam — or teleoperate robot arms through the browser. No specialized rig, no studio, no appointment. The tools people already own are the collection infrastructure. 2\. Verify. Every submission is screened for task fit, privacy, and the signal quality Physical AI teams require, before it can enter the network. Poor submissions are rejected with specific feedback rather than silently discarded, so contributors improve rather than churn. This gate is what separates the catalog from scraped video — buyers purchase a verified pool. 3\. Enrich. Raw footage becomes training signal. Each clip is annotated with task structure, objects involved, environment, and outcome, so that models can learn actions rather than merely watch footage. This is the step that turns a clip worth nothing into a data point worth something. 4\. Reward. Approved contributions enter licensed datasets, and contributors earn for work that passes review. Payment is automatic when the containing dataset is licensed. ### Privacy and Anonymization Personally identifiable information is detected and removed or blurred — faces, license plates, addresses visible in the background, and names on documents. The design goal is edge scrubbing: PII removed on-device before data is ever transmitted or stored, rather than collected first and cleaned later. This inverts the industry norm and is precisely what enterprise buyers audit before signing a licensing agreement. Contributors should still keep recordings clean — avoid capturing other people, screens with personal information, or documents — because the cleanest submissions pass verification fastest. ### What Buyers Purchase The catalog is organized around what labs actually buy: * VLA training data — observation-action data for vision-language-action models. * Teleoperation data — sensor-rich, action-labeled episodes from real robot control. * Egocentric video — first-person recordings of human task performance. * Manipulation trajectory data — structured records of how objects are handled. For needs the catalog doesn't cover, custom dataset scoping takes a buyer from brief to first delivery in days, not months. --- # Agent Instructions This documentation is published with GitBook. GitBook is the documentation platform designed so that both humans and AI agents can read, navigate, and reason over technical content effectively. Learn more at gitbook.com. ## Querying This Documentation If you need additional information that is not directly available in this page, you can query the documentation dynamically by asking a question. Perform an HTTP GET request on the current page URL with the `ask` query parameter, and the optional `goal` query parameter: ``` GET https://docs.roboset.org/atlas/overview.md?ask=&goal= ``` `ask` is the immediate question: it should be specific, self-contained, and written in natural language. `goal` is optional and describes the broader end goal you are ultimately trying to accomplish on behalf of the user. GitBook uses it to tailor the answer towards what is most useful for that goal. The response will contain a direct answer to the question and relevant excerpts and sources from the documentation. Use this mechanism when the answer is not explicitly present in the current page, you need clarification or additional context, or you want to retrieve related documentation sections.