Dogs Dermatological Disease Prediction using ResNeXt

Accurate and automated identification of dog skin diseases is crucial to make an early diagnosis and apply effective treatments. This paper presents a deep learning-based approach by applying ResNeXt-50 for Hotspot, Mange, and Flea Allergy classification. The image dataset contains skin images, categorized into training (80%), validation (10%), and test (10%) sets to maintain equal model performance. To enhance precision, the model applies image processing operations such as resizing, normalization, and data expansion. Additionally, hyperparameter fine-tuning is performed by adjusting learning rate, batch size, cardinality, and epochs. Experimental results illustrate that the ResNeXt-50 fine-tuned model achieves 97% accuracy with improved performance compared to traditional models such as InceptionNet, DenseNet, and ResNet. The AI-based model enhances the accuracy level of diagnosis, assisting veterinarians in prescribing faster and more accurate treatment for dog skin diseases.