Multi-Livestock Disease Classification using Deep Learning: A Comparative Study of Pretrained CNN Architectures

The precise and quick identification of livestock illnesses functions as the essential requirement which maintains worldwide food security while enabling environmentally friendly farming methods to proceed. The health of livestock establishes the foundation which enables farmers to achieve consistent food production while decreasing their economic losses. The process of early and dependable disease detection allows medical professionals to execute their work effectively which results in better animal treatment and controls disease transmission between animals. The research paper examines the performance of eight different classification models through comparative analysis as they detect multiple diseases that affect livestock. The models include four pretrained convolutional neural networks (CNNs) which are EfficientNet- B3 DenseNet121 ResNet50 and MobileNetV2 together with additional baseline approaches used to evaluate performance. The study uses a practical dataset which contains 19621 images that belong to 19 different disease classes which affect four main livestock categories, including Cow, Goat, Pig, and Poultry. The researchers applied complete data augmentation methods together with transfer learning techniques to enhance the models’ capacity to handle different situations. The experimental results show that EfficientNet-B3 delivered the top performance with an accuracy of 91.4% and an F1-score of 0.912. The model surpassed traditional baseline models by more than 20 percentage points. The research demonstrates that compound-scaled pretrained CNN architectures deliver dependable automated solutions for detecting livestock diseases across multiple species in contemporary agricultural environments.