Research note: Application of convolutional neural networks for feather classification in chickens

Feather color plays a crucial role in distinguishing various poultry breeds. However, the reliability of conventional manual classification methods is debated due to the intricate nature of feather color traits. To address this issue, we applied Convolutional Neural Networks (CNN) to extract the fea...

Full description

Saved in:
Bibliographic Details
Main Authors: Jiajia Niu, Tong Li, Kunlong Qi, Yang Liu, Haixuan Deng, Yujia Hu, Dan Xu, Liuting Wu, Felix Kwame Amevor, Yingjie Wang, Gang Shu, Xiaoling Zhao
Format: Article
Language:English
Published: Elsevier 2025-11-01
Series:Poultry Science
Subjects:
Online Access:http://www.sciencedirect.com/science/article/pii/S0032579125004961
Tags: Add Tag
No Tags, Be the first to tag this record!
Description
Summary:Feather color plays a crucial role in distinguishing various poultry breeds. However, the reliability of conventional manual classification methods is debated due to the intricate nature of feather color traits. To address this issue, we applied Convolutional Neural Networks (CNN) to extract the feather texture features of 300 images each of Golden Meihua (GM) and Silver Meihua (SM) feathers. The nonlinear features were then learned through an MLP layer and activation functions to complete the classification of GM and SM. Finally, we successfully developed a method automating the identification of feather texture features, achieving a recognition model accuracy of 93.71 % after 5-fold cross-validation. This research improves the precision and effectiveness of feather color selection, offering valuable insights into the systematic classification of feather colors in diverse poultry species.
ISSN:0032-5791