French researchers build AI to monitor sheep behavior via collars
Sheep wear AI-powered collars to track grazing, rumination and heat stress in real-time
EweAcT is a new dataset linking neck-collar accelerometer data to sheep behavior, built from 79 hours of tri-axial sensor recordings and video-annotated observations of 120 Romane ewes. The animals were raised in an extensive outdoor system on 280 hectares of rangeland, with data collected under real-world conditions like sloping pastures and heat waves. Behaviors include grazing, ruminating, resting, moving, and other activities. The dataset is designed to be ready for artificial intelligence models to classify the main behaviors of sheep in extensive grazing systems.
- EweAcT dataset contains 79 hours of aligned accelerometer and video data from 120 Romane sheep wearing AI-powered collars
- Behaviors include grazing, rumination, resting, moving, and others, annotated using Behavioral Observation Research Interactive Software
- Models can predict sheep well-being and stress responses in real-time across 280 hectares of rangeland
Why It Matters
AI-driven livestock monitoring can improve animal welfare and help farmers adapt to climate change and environmental stressors.