Why it matters
Without this system, misread signals can lead to escalating stress, injuries, repeated reactive cases, inconsistent welfare assessments, and uncertain certifications or insurance claims.
From design to safety
Design
AI-informed behaviour space design
Defines data flows from video capture to automated alerting, framing threshold criteria and welfare-first objectives for the consultation rooms and reactive spaces; includes data dictionary, retention timelines, and ethics review.
- lip licking observation station
- whale eye camera rig
- tail carriage monitor
Build
Install sensing and analytics pipeline
Sets up cameras, sensors, and data stores; configures privacy, data retention, and integration with case management and risk registers; validates data pathways with test dogs and dummy sessions.
- freezing sensor mat
- stress score analytics workstation
- behavioural data server
Operation
Real-time detection and triage
Processes live video, extracts cue cues, updates stress scores, and triggers supervisory review when thresholds are crossed; provides annotated clips and incident notes for case files.
- lip licking observation station
- whale eye camera rig
- tail carriage monitor
Maintenance
Ongoing calibration and quality checks
Regularly tests sensor accuracy, retunes AI thresholds, validates annotations against canine welfare plans, updates training materials, and archives drift reports for QA.
- freezing sensor mat
- stress score analytics workstation
- behavioural data server
Safety
Welfare-first decision support
Ensures only qualified staff sign off on interventions; links to CPD records, insurance checks, and professional standards auditing; enforces escalation to veterinary advice where needed.
- lip licking observation station
- whale eye camera rig
- tail carriage monitor
Part by part
- 1
lip licking observation station
- Watch for
- lip licking frequency spikes and associated sniffing, indicating arousal or appeasement; detect pacing paired with lip actions.
- AI action
- AI flags potential stress cue clusters and proposes a welfare review note.
- 2
whale eye camera rig
- Watch for
- eye contact avoidance, scleral exposure, and gaze shifts indicating threat perception or fear.
- AI action
- AI computes stress score delta and highlights frames for clinician review.
- 3
freezing sensor mat
- Watch for
- sudden immobility lasting beyond threshold duration, suggesting high anxiety or risk of withdrawal.
- AI action
- AI raises an alert and logs freezing episode with timestamp and context.
- 4
tail carriage monitor
- Watch for
- tail position changes and stiffness signaling arousal or tension.
- AI action
- AI marks event as potential stress marker and suggests step-down or de-escalation plan.
- 5
stress score analytics workstation
- Watch for
- aggregate stress scores across sessions; detect trend toward elevated baseline.
- AI action
- AI generates a case-level stress trajectory report for the behaviour consultant.
- 6
behavioural data server
- Watch for
- data integrity issues, missing annotations, or inconsistent scoring across examiners.
- AI action
- AI flags data quality anomalies and recommends review by the lead consultant.
Worked example
Detecting rising stress in a reactive dog during a board-and-train intro
Illustrative scenario with sample data.
A reactive border collie shows lip licking, whale eye, and occasional freezing during initial exposure in the consultation room.
Signal in
videoFrames:30fps, dogID:BC-07, sessionID:S-101, timestamp:2026-11-08T09:00:00Z, cues:lipLicking=2.1/s, whaleEye=0.9, freezing=0.3s, tail=neutral
System out
Alert: elevated stress score trend, frames highlighted for clinician review; recommended welfare check and short pause with plan for gradual exposure progression.
Human sign-off: Behaviour consultant reviews highlighted frames, confirms welfare concern, adjusts exposure, documents plan in record, and notifies trainer team of next steps.
How to roll it out
Phase 1
Weeks 1 to 4: Baseline
Install sensors, calibrate thresholds, and establish data flows with initial staff training.
Phase 2
Weeks 5 to 8: Pilot run
Pilot with 3 dogs, test alerting accuracy, and refine scoring against welfare plans.
Phase 3
Weeks 9 to 12: Full integration
Roll out across reactive spaces, integrate with CPD tracking and case management.
Where people stay in charge
- Welfare interventions must be signed off by a certified behaviour consultant or veterinary behaviourist when medical questions arise.
- Data sharing with external parties requires academy director approval and client consent in records.
- No automation should override professional judgement in interpreting stress cues or distinguishing fear from excitement; human oversight remains mandatory.
Questions owners and crews ask
What data feeds are used for AI body language detection in this system?
The system reads video frames from the behaviour consultation rooms, sensor readings from the freezing mat, and annotator inputs to compute moment-by-moment cues and stress scores.
Who approves welfare-related actions suggested by the AI?
A certified behaviour consultant signs off on welfare interventions, with a veterinary advisor consulted for medical questions.
How are privacy and consent handled in behaviour spaces?
Camera use is disclosed in client agreements; data retention complies with local policies and consent is obtained for recordings used in training.
Can the system track long-term progress for a case?
Yes, the stress score trajectory and cue response history feed into the Canine Good Citizen progression and modification plans.