Project Objectives

• Install a complete kennel‑wide video‑recording infrastructure, including infrared lighting for night‑time monitoring.
• Develop and train multi‑animal tracking networks capable of identifying individual dogs and predicting key body points.
• Build personalised behaviour profiles for each dog using behaviour‑flow and activity‑pattern analysis.
• Compare machine‑learning–based behavioural assessments with current gold‑standard welfare monitoring.
• Evaluate behavioural changes in response to interventions (training, enrichment, environmental changes, stressors).
• Create a mobile recording app for synchronised multi‑camera data collection.
• Share tools and insights with other research groups and support broader cross‑species adoption.

How This Advances 3Rs Implementation

• Replaces subjective welfare monitoring with objective, data‑driven methods.
• Improves welfare by enabling undisturbed 24/7 observation, capturing subtle behavioural shifts.
• Supports reduction by increasing the richness and quantity of data per animal.
• Allows cross‑facility implementation through protocol sharing and training.
• Improves reproducibility by reducing variation introduced by human observers.

Background

Behavioural monitoring is central to ensuring high welfare standards in research animals, yet current approaches for dogs often rely on qualitative observations or score sheets, that are subjective and insensitive to subtle changes. While advanced, automated behavioural analysis is increasingly used in rodents, these technologies have not yet been widely applied to dogs. This leaves a knowledge gap in objective, continuous monitoring methods for canine welfare, particularly in kennel environments where behavioural changes may indicate stress, discomfort, or environmental needs.

This project set out to translate modern machine learning based tracking tools to dogs, establishing a system capable of continuously recording and analysing behaviour in both indoor and outdoor kennel spaces. The team upgraded the facility with video‑recording infrastructure, including infrared lighting for night monitoring and customised enrichment objects detectable by the computer‑vision models. Through multi‑animal tracking, pose estimation, and behaviour‑flow analysis, the project generated individual behaviour profiles that allow welfare assessments tailored to each dog.