Research Dashboard Overview
The Research Dashboard turns recordings and analysis results into a review workspace. Every screenshot below is clearly generated from Simulated Research Data.
What the words mean
- A session is one recording plus its synchronized metadata and sensors.
- An analysis run is one immutable set of derived results made with a recorded configuration and model version.
- A bounding box marks where a model thinks an object appears. It can be wrong.
- A track links detections believed to represent the same object across frames.
- Confidence is the model’s score, not a guarantee of correctness.
- Natural, manufactured, and unknown are normalized experimental evidence scores with explanations. They are not scientific ground truth.
- A warning describes something worth knowing about capture, storage, sensors, networking, analysis, compatibility, integrity, temperature, import, or export.
Human review
Labels, boxes, tracks, and scores require human review. Reviewers can mark results verified, incorrect, duplicate, ambiguous, ignored, or useful as a training candidate. Phase 6 does not add freehand box drawing or final training-data export.
Device roles
- iPhone: records sessions, shows live detection and connection status, and supports lightweight inspection.
- iPad: provides portable live monitoring, session review, charts, filters, bookmarks, warnings, and run comparison.
- Mac: provides the largest catalog, chart, object-index, track-detail, comparison, and batch-review surfaces.
