A workspace organised around the study.

Labs provide the collaboration boundary; projects hold the recordings, configuration and results. The workspace brings the stages of a study into one navigable record.

  • Sources & setup

    Upload recordings, register experimental metadata and define timestamps and apparatus regions.

  • Pose estimation

    Submit inference and fine-tuning jobs to hosted GPUs; inspect their progress and resulting artefacts.

  • Annotation & review

    Claim review tasks, publish reviewed annotations and retain corrections as explicit inputs.

  • Analysis

    Compare configuration versions and runs, then inspect individual trajectories and cohort summaries.

  • Quality & validation

    Examine quality reports alongside the recording and the assumptions of the analysis.

  • Exports

    Collect results, configuration and provenance into a research bundle.

Beginning a project.

  1. Establish the context. Create or open a lab, then create a project for the study.
  2. Register the source. Upload a recording and identify the animal, cohort and experimental phase.
  3. Define the apparatus. Mark the relevant regions and check the timestamp segments.
  4. Run and inspect. Launch the appropriate jobs, examine their outputs and review uncertain observations.
  5. Validate and export. Retain the report and configuration alongside the data used in the study.
Open the workspace

Practical questions.

Do I need a local GPU?

Pose estimation and fine-tuning are dispatched to hosted hardware. The research workspace is accessed through the browser.

Which experiments is Fylgja designed around?

The present workflow centres on mouse behaviour, particularly the three-chamber social interaction assay. Suitability for a different species, camera arrangement or assay requires separate assessment.

Can I inspect an earlier result?

The workspace includes job history, analysis runs, configuration versions and provenance. Use those records to establish which inputs and settings produced the result.

Does a quality report establish biological validity?

A quality report describes checks on a particular run. Behavioural validity is established separately, against independent human assessment. See Evidence.