Feline Feelings Unleashed: Harnessing Deep Learning Through Photos for Cat Pain Detection

Abstract

Effective pain detection in cats is challenging. Thus, we classified feline pain using a dataset of 57 images per category, labeled 'pain' or 'no pain' by Thai veterinarians using the performance of EfficientNetB7, a pre-trained convolutional neural network. The images were preprocessed through various configurations of EfficientNetB7 in differing batch sizes and learning rates. ImageNet weights were used as the initial training parameters. The models were evaluated based on accuracy, precision, and recall. The most effective model was the SGD optimizer with a learning rate of 0.001 and a batch size of 100. The model achieved 79% accuracy, 74% precision, and 90% recall. These findings demonstrated the potential of deep learning for non-verbal pain detection in veterinary, especially with the high recall rate to identify animals in distress. The research results showed the possibility of integrating such AI models into veterinary practice, enhancing animal welfare.