Lumpy Skin Disease Detection Using VGG19 and ANN Hybrid Model

Abstract

Lumpy Skin Disease (LSD) is a cattle disease caused by a virus and highly contagious and as a result the livestock sector incurs a huge financial loss across the globe. Outbreaks of LSD must be contained and the mortality rate must be kept low, therefore early LSD diagnosis is essential. Traditional diagnosis methods, however, are slow, need a specific type of expertise, and in rural areas the diagnostics are usually out of reach. These factors contribute to a delayed outbreak response and as a result increase the financial strain on the farmers. New disease detection methods that utilize automation and artificial intelligence can provide viable alternatives. In this paper, we present a novel methodology in which we fuse both deep learning and machine learning methods in order to detect LSD accurately. We utilize the VGG19 Convolutional Neural Network (CNN) for the classification of images of cattle with skin disease, and a skin disease detection artificial neural network (ANN) model for disease detection, trained on CSV clinical and environmental features. The disease prediction related image and tabular data used in this study are obtained from Kaggle. The VGG19 model is fine-tuned with the use of transfer learning to develop its feature extraction and classification components. The ANN model has been trained using normalized input features and has been optimized through the use of backpropagation. The amalgamation of these two models increases the stability and reliability of the detection system. Results from experiments show that the VGG19 model demonstrates a record of 92.16% with a loss of 0.2138. On the other hand, the ANN model also shows record of higher accuracy with 96% and the data is said to be structured. In its 0.9215 of the validation accuracy, the VGG19 model shows good performance of ability to generalize. The detection accuracy is further enhanced with the use of the suggested combination of models and this along with its flexibility and simplicity and low cost overall, aids in the early detection of the disease called LSD. With this system in place, it aids in the rapid and effective decision making of the farmers along with the vets and this in turn aids to quell the spread of the disease and improves the overall management of the livestock in the health aspect.