Intelligent Poultry Health Recognition from Fecal Images Using YOLO11n

Abstract

: Poultry industry plays a vital role in the livestock industry. In poultry production, disease breakouts can cause a fall in production performance and, at times, can cause significant mortality which could result in tremendous economic loss. To realize the low-cost, contactless poultry health status detection for large-scale poultry houses, this study explores the method of detecting the four types of poultry health status according to the images of poultry feces. A total of 6812 photographs were taken on farms, of four types: Healthy, Coccidiosis, Newcastle Disease and Salmonellosis. The dataset has been divided into training, test and validation sets in a ratio of 7:2:1. In this study YOLO11n models were trained from scratch and with pretrained weights. The results show that the pretrained model achieved Precision, Recall, mAP@0.5 and mAP@0.5:0.95 of 85.1%, 73.1%, 83.6%, and 63.1%, respectively; the model trained from scratch achieved 81.5%, 76.0%, 82.8%, and 59.9%, respectively. In addition, pretraining boosted the overall Precision by 3.6%, mAP@0.5 by 0.8%, and mAP@0.5:0.95 by 3.2%. This procedure can be used as an automatic preliminary screening procedure for abnormal feces in chicken houses.