Related Experiment Video
Updated: Aug 22, 2025
06:19
Integration of Animal Behavioral Assessment and Convolutional Neural Network to Study Wasabi-Alcohol Taste-Smell Interaction
Published on: August 16, 2024
485
An Advanced Chicken Face Detection Network Based on GAN and MAE.
Xiaoxiao Ma1, Xinai Lu2, Yihong Huang3
1College of Information and Electrical Engineering, China Agricultural University, Beijing 100083, China.
Animals : an Open Access Journal From MDPI
|November 11, 2022
Summary
This study introduces a new chicken face detection system using generative adversarial network-masked autoencoders (GAN-MAE) data augmentation. The enhanced model achieves higher accuracy and speed for smart poultry farming applications.
Area of Science:
- Computer Vision
- Artificial Intelligence
- Agricultural Technology
Background:
- Accurate chicken face detection is crucial for smart poultry agriculture but faces challenges with limited datasets and low-accuracy models.
- Existing algorithms struggle with small object detection, hindering precision management in large-scale farms.
Purpose of the Study:
- To develop an advanced object detection network for high-accuracy chicken face detection across different ages.
- To address the scarcity of chicken face datasets and improve detection speed and accuracy.
Main Methods:
- Utilized generative adversarial network (GAN) and masked autoencoders (MAE) for data augmentation.
- Employed CSPDarknet53 as the backbone network and improved feature fusion with dense connections for YOLO head.
- Modified feature map downsampling to capture smaller object features.
Main Results:
- Achieved a mean average Precision (mAP) of 0.84, a 29.2% improvement over existing networks.
- Maintained a detection speed of 37 frames per second.
- Demonstrated superior detection accuracy meeting practical requirements.
Conclusions:
- The proposed GAN-MAE based object detection network significantly enhances chicken face detection accuracy and efficiency.
- The developed system provides a practical solution for smart poultry agriculture, enabling precision management.
- The end-to-end web system facilitates the real-world application of this advanced algorithm.
