Related Experiment Video
Updated: May 14, 2026
06:37
Artificial Intelligence-Based System for Detecting Attention Levels in Students
Published on: December 15, 2023
Detecting Dairy Cattle Protective Behaviors via a Multi-Stage Attention SlowFast Network.
Bo Zhang1, Jia Li1,2, Feilong Kang3
1College of Computer and Information Engineering, Inner Mongolia Agricultural University, Hohhot 010018, China.
Animals : an Open Access Journal From MDPI
|May 13, 2026
Summary
This study introduces a new AI model, MSA-SlowFast, to accurately detect dairy cattle protective behaviors for better pasture management. The model significantly improves detection accuracy, aiding in animal welfare monitoring.
Area of Science:
- Animal Science
- Computer Vision
- Artificial Intelligence
Background:
- Protective behaviors in dairy cattle are key indicators of health and welfare.
- Current detection methods struggle with rapid movements, background noise, and imbalanced data in agricultural settings.
Purpose of the Study:
- To develop and evaluate a novel AI model for precise detection of dairy cattle protective behaviors.
- To improve pasture management and animal welfare through enhanced behavior monitoring.
Main Methods:
- Proposed the Multi-Stage Attention SlowFast (MSA-SlowFast) model, an enhanced SlowFast network.
- Incorporated Multi-Path Balanced Head (MPBHead), Spatio-Temporal Convolutional Block Attention Module (ST-CBAM), and 7 (BAF) modules.
- Developed timing-aware oversampling and dynamic loss adjustment for minority class improvement.
- Constructed a spatio-temporal dairy cattle protective behaviors dataset.
Main Results:
- The MSA-SlowFast model achieved 79.41% mAP, outperforming standard SlowFast (70.58%) and Slow-only (68.21%).
- High detection confidence was observed for specific protective actions: tail swaying (0.97), head shaking (0.90), ear flapping (0.92), and leg kicking (0.90).
Conclusions:
- The MSA-SlowFast model demonstrates significant feasibility and value for detecting dairy cattle protective behaviors.
- The developed methods offer a promising approach for improving animal welfare monitoring in complex environments.