Workplace Group Stress Heatmap

Problem Context

In open-plan offices, team stress often appears as localized tension, rigid posture, eye rubbing, or forward leaning. These signals are spread across workstations, making it hard to judge zone-level load from isolated observations. Individual identification would cross privacy and compliance boundaries.

How It Works

The skill takes fixed-camera office video as input, supporting mp4, avi, and mov files up to 10MB, preferably from a high or oblique view covering multiple workstations. It aggregates workstation ROI signals such as:
- Facial tension: frown intensity and downturned mouth corners
- Posture rigidity: duration of unchanged pose and shoulder elevation angle
- Fatigue-related events: eye-rubbing counts and forward-leaning proportions

It then computes zone-level stress indices with a group assessment model and produces a green-to-red pseudo-color heatmap. If a zone has fewer than 3 people, the result is marked insufficient_sample instead of emitting an individual-inferable stress_index. Historical reports are queried from the cloud API and usually rendered as Markdown tables with report links.

Boundaries

It fits open-plan offices, R&D centers, and customer-service centers where team-level state monitoring is needed. It is not for facial recognition, performance evaluation, or individual behavior tracking. Before deployment, disclose collection scope, retention, and purpose, and avoid interpreting discussion, reading, or thinking directly as stress.