Stacking Ensemble Deep Learning for Automated Broiler Breeder Feather Scoring
Stephen Afrifa, Sawyer Exum, Biswash Khatiwada, Kapalik Khanal, Edgar Orlando Oviedo-Rondon, Ramesh Bahadur Bist
North Carolina State University Tribhuvan University
阅读操作
确认中在文库中上传 PDF 后可生成中文音频讲解。
摘要与影响
Assessing feather condition in broiler breeders is essential for monitoring poultry welfare and identifying feather damage and loss associated with management, health, natural mating, and behavioral challenges. Feather condition can be associated with fertility failures, as featherless areas can suffer scratches from roosters during mating, reducing reproductive success. Manual feather scoring (FS) is labor-intensive, subjective, and difficult to scale in broiler breeder production. This study proposed a stacking ensemble framework for automated multiclass FS classification using RGB images. A total of 67,231 images were categorized into six FS classes (FS_0–FS_5) and used to train three independently trained models: MobileNetV2, EfficientNetB0, and ResNet50. Their class probability outputs were integrated using a Logistic Regression (LR) meta-learner. The proposed stacking ensemble achieved 99.83% accuracy, 99.83% macro precision, 99.86% macro recall, 99.84% macro F1-score, an Area Under the Receiver Operating Characteristic Curve (AUC) of 0.9999, and a quadratic weighted kappa (QWK) of 0.9997. The QWK and ordinal error analysis indicated near-perfect chance-corrected agreement, with no severe cross-category misclassifications. Bird-grouped five-fold cross-validation showed stable performance within the experimental domain evaluated. Gradient-Weighted Class Activation Mapping (Grad-CAM) provided qualitative insight into model attention while also indicating some background activation. The proposed framework shows strong performance on a finer six-class FS task and demonstrates the potential of stacking ensemble learning for objective poultry welfare assessment. Broader cross-farm and cross-device validation remains necessary before practical deployment.
逐年被引趋势
暂无年度引用数据
关键指标
同类平均 = 1
同领域 · 同年份 · 同类型
Google Scholar 与 OpenAlex 的被引统计范围不同,数值存在差异属正常。
AI 辅助阅读
依据:摘要
可就本文提问;依据不足时会说明。
学术脉络
学科主题
生物医学Smart Agriculture and AI
Advanced Neural Network Applications · Animal Behavior and Welfare Studies
参考文献 34
此处列出前 3 条