Image-Based Camel Skin Diseases Diagnosis Using Convolutional Neural Network

Abstract

This study was conducted on image-based camel skin diseases using Neural Network. In developing countries such as Ethiopia, livestock plays a great role as socioeconomic assets. Among that livestock camel is the most vital domestic animal species for pastoralist livelihood. The basic role of the camel is to serve humans directly in extremely harsh conditions. However, due to several factors, camel’s contributions to the human welfare of developing countries such as Ethiopia are generally obscured. Among those factors, diseases are number one. Among those diseases, skin diseases are the ones that affect Camel in their life as well as minimize their productivity. There are a lot of research works that have been conducted in the area of human skin diseases classification but only a few works conducted in the area of animal skin diseases classification. Especially, to the knowledge of the researcher, there is no research conducted on camel skin diseases classification using machine learning approach. Therefore, the objective of researcher was to investigate and developed a model that diagnoses a camel skin disease from the image using Convolutional Neural Network. A three-class identification of camel skin diseases has been envisioned, developed, and accomplished in this thesis. For this study, the researcher collected 3,682 images directly from affected camel skin and 34 images from public sources. A Convolutional Neural Network was utilized for feature extraction from deep learning, and the SoftMax activation function was used for classification. Keras with TensorFlow framework was used in this study along with Visual Geometry Group Network architecture for transfer learning due to its speed. Seven experiments were conducted to develop a model. The second experiment resulted in an accuracy of 93.41%, while transfer learning achieved an accuracy of 98.21%. The researcher suggested that further investigation into different camel skin diseases be conducted for diagnostic purposes.