Related Experiment Video
Updated: Apr 30, 2026
05:41
A Step-by-Step Implementation of DeepBehavior, Deep Learning Toolbox for Automated Behavior Analysis
Published on: February 6, 2020
8.7K
Monitoring horse behaviour with deep learning models.
Claudia Giannone1, Chiara Maccario2, Emanuela Dalla Costa2
1Department of Agricultural and Food Sciences, University of Bologna, Bologna, Italy.
The Veterinary Quarterly
|April 28, 2026
Summary
Deep learning accurately tracks stabled horse behaviors like standing and lying using video analysis. This non-invasive method aids in automated equine welfare monitoring.
Area of Science:
- Animal behavior
- Machine learning
- Equine science
Background:
- Assessing stabled horse well-being is crucial.
- Non-invasive methods for behavior analysis are advancing.
- Deep learning offers potential for automated monitoring.
Purpose of the Study:
- To evaluate a convolutional neural network (CNN) for recognizing equine behaviors.
- To assess the accuracy of deep learning in classifying standing, lying, and drinking in horses.
- To analyze activity patterns using video data.
Main Methods:
- A CNN model was trained and evaluated using video data of a horse over 29 days.
- Continuous video recording in a wooden stall.
- Model predictions were compared against manually annotated ground truth data.
Main Results:
- High precision (97.5%) and recall (89.2%) for detecting standing behavior.
- High precision (92.8%) but lower recall (63.1%) for lying behavior.
- Standing comprised over 85% of daily activity; lying accounted for 5-10%; drinking peaked between 4 PM and 10 PM.
Conclusions:
- Deep learning models can effectively classify common equine behaviors from video.
- This technology supports automated monitoring of horse welfare.
- Further research is needed to recognize less frequent behaviors.
