Related Experiment Video
Updated: May 3, 2026
05:41
A Step-by-Step Implementation of DeepBehavior, Deep Learning Toolbox for Automated Behavior Analysis
Published on: February 6, 2020
9.4K
GFI-YOLOv8: Sika Deer Posture Recognition Target Detection Method Based on YOLOv8.
He Gong1,2,3, Jingyi Liu1, Zhipeng Li4
1College of Information Technology, Jilin Agricultural University, Changchun 130118, China.
Animals : an Open Access Journal From MDPI
|September 28, 2024
Summary
This study introduces GFI-YOLOv8, an advanced algorithm for sika deer posture recognition, improving animal welfare monitoring. The model achieves 91.6% accuracy, enhancing health assessments in large-scale breeding operations.
Area of Science:
- Computer Vision
- Animal Science
- Artificial Intelligence
Background:
- Sika deer farming requires accurate health monitoring for welfare and productivity.
- Traditional monitoring methods can be invasive and may disrupt natural animal behavior.
- Posture recognition via target detection offers a non-intrusive approach to assess animal well-being.
Purpose of the Study:
- To develop a novel algorithm for remote, non-invasive sika deer posture recognition.
- To enhance the accuracy and efficiency of monitoring sika deer health and behavior.
- To improve animal welfare standards in large-scale sika deer breeding operations.
Main Methods:
- Proposed GFI-YOLOv8 algorithm, an enhancement of YOLOv8 for sika deer posture detection.
- Incorporated iAFF iterative attention feature fusion module and AIFI module.
- Introduced a novel convolutional neural network module, attention mechanism, pyramid network, and optimized detection head.
Main Results:
- GFI-YOLOv8 achieved 91.6% accuracy in sika deer posture recognition.
- Demonstrated a 6% improvement in accuracy and 4.6% increase in mAP50 over YOLOv8n.
- Outperformed other YOLO series models (YOLOv5n, YOLOv7-tiny, YOLOv8n, YOLOv8s, YOLOv9, YOLOv10) in key metrics.
Conclusions:
- The GFI-YOLOv8 algorithm provides accurate and rapid identification of sika deer postures.
- The model is suitable for real-time monitoring in complex, all-weather breeding environments.
- This technology significantly contributes to improved animal welfare and management in the sika deer industry.
