← Data Egocentric

Fiund / Dataset solution

Egocentric data for temporal reasoning and memory

Source contextual first-person sequences for event retrieval, ordering, and evidence-based answers.

Questions about what happened earlier depend on footage that preserves context and reliable references to time. Define the event your model should find or explain, the available history, and the evidence expected in an answer. A collection designed for short action recognition may omit the context this task requires.

Choose answerable questions

Use questions tied to observable events, such as where an object was placed or which step preceded cleanup. Separate visible facts from intentions and information outside the recording. Write examples of valid answers, evidence spans, and questions that should receive an insufficient-evidence response.

Retain the relevant history

Decide whether models receive continuous sessions, linked episodes, or a limited window. Cutting around a single interaction may remove the earlier object placement that makes a later question answerable. Preserve source-session identifiers and original time references when creating derived clips or query annotations.

Review ambiguous references

Repeated objects and repeated actions can make a question underspecified. Explain whether references identify an object category, a particular instance, or the most recent event. Ask reviewers to check whether two reasonable answers exist and to flag cases in which an object disappears from view before its destination is known.

Protect evaluation independence

Keep related footage and derived questions together when dividing training and evaluation. Otherwise a model may encounter nearly the same event through a different crop or narration. Record the conditions held out, and inspect evidence spans independently of answer wording to catch fluent answers unsupported by the video. If several similar events occur, require evidence that identifies the intended occurrence rather than an interchangeable-looking moment elsewhere in the session.

Include in your inquiry

  • Question types and evidence format
  • Available history and session structure
  • Ambiguity and unanswerable-question rules
  • Session grouping for evaluation

Your project, in focus

Describe the first-person view your model needs.

Speak directly with Jaeden Schafer at Fiund about the source material, specifications, and review your project needs.

Source fit, permissions, samples, pricing, and delivery are reviewed for the individual project.

Start with these details

  1. Tasks and environments
  2. Camera placement and visibility
  3. Target hours or sequences
  4. Annotations and intended model use
  5. Rights requirements and timeline
Build your project brief ↗