Classifying Dog Emotions Using Transfer Learning and Pre-trained Models

Dogs’ domestication fosters close bonds with humans, enabling them to feel emotions like joy, sorrow, and frustration. To provide dogs with adequate care, it is essential to understand their emotional states. Previous research has explored deep learning techniques for emotion recognition in dogs. However, there is a need for improved accuracy and more efficient computational models. This paper proposes a novel neural architecture to predict dog emotions with optimal accuracy. The approach builds on top of three existing pre-trained Convolutional Neural Network models and uses a transfer learning approach to fine-tune these models for dog emotion prediction. The methodology and the novel architecture achieved a high accuracy of 93%. The proposed architecture works exceptionally well, providing a foundation for a system that veterinary professionals can use to understand their dogs' emotions. Overall, this study provides a novel strategy for predicting dog emotions and offers a promising direction for future research.