Enhanced Model for Chicken Disease Detection Improves Food Safety

Strategic Insights -

A team of researchers from Shanxi Agricultural University in China has developed an advanced artificial intelligence model, DOSA-YOLO, to detect four common chicken diseases, aiming to improve the global food supply's sustainability. This model is designed for real-time chicken health monitoring in intensive farming environments. The researchers utilized a combination of phenotypic features, such as comb, eyes, and wattles, as well as pathological anatomical characteristics to develop a five-class dataset comprising 8052 images.

Key Takeaways:

  • The researchers developed an enhanced YOLOv11 model, DOSA-YOLO, to detect avian pox, coccidiosis, *Mycoplasma gallisepticum*, and Newcastle disease in chicken farms.
  • The model was trained and validated using a constructed dataset of 8052 images, categorized based on phenotypic features and pathological anatomical characteristics.
  • Three attention-enhancement modules, MSDA, MDJA, and SEAM, were integrated into the YOLOv11 model to address complex backgrounds, multi-scale lesions, and occlusion interference.
  • The model achieved a mean Average Precision (mAP) of 97.2% and an F1-score of 95.0%, outperforming seven other algorithms, including YOLOv5n, YOLOv7tiny, YOLOv8n, YOLOv9t, YOLOv10n, YOLOv11n, and YOLOv12n.
  • The model provides strong support for real-time chicken health monitoring in intensive farming environments, enhancing the global food supply's sustainability.
  • The research team included Xiaofeng Guo, Yun Wang, Jianhui Li, Qin Li, Zhenhuan Zuo, and Zhenyu Liu, with support from The Key R&D Program of Shanxi Province and Shanxi Scholarship Council of China.
  • The model maintained a balance between lightweight design and performance, with GFLOPs of 6.9 and 2.87 M parameters.

Statistics:

  • 8052 images were used in the dataset to train and validate the model.
  • The model was trained and validated using a combination of phenotypic features and pathological anatomical characteristics.
  • The model achieved a mean Average Precision (mAP) of 97.2% and an F1-score of 95.0%.
  • GFLOPs of 6.9 and 2.87 M parameters maintained a balance between lightweight design and performance.

Sources:

  • DOSA-YOLO: Improved Model Research for the Detection of Common Chicken Diseases Using Phenotypic Features. Agriculture, 2025, 15(19):1996. (Agriculture - http://www.mdpi.com/journal/agriculture)
  • Xiaofeng Guo, Yun Wang, Jianhui Li, Qin Li, Zhenhuan Zuo, and Zhenyu Liu. DOSA-YOLO: Improved Model Research for the Detection of Common Chicken Diseases Using Phenotypic Features. Agriculture, 2025, 15(19):1996.