Intelligent Application for Dog Emotion Detection using EfficientNet-B0

The present research paper introduces an innovative methodology for classifying canine emotions by employing the EfficientNet-B0 algorithm. The recognition of emotions by animals, specifically canines, is a crucial factor in comprehending their conduct and improving the quality of human-animal relationships. This research study addresses the task of precise dog emotion recognition, like happy, sad, angry, relaxed. In this study, the model, fine-tuned, is used to cover different emotions in dogs and has a classification accuracy of 89%, thus showing its potential to recognize emotions like anger, happiness, relaxation and sadness among dogs. The Efficient net- B0 model is meticulously trained on an augmented dataset known as Stanford’s Dog Emotion dataset, which consists of images of dogs with labels indicating their emotional state: angry, happy, relaxed, or sad. Metrics that show constant performance, such as accuracy, loss, AUC, and precision when viewed visually, show consistent performance. According to the data, EfficientNet-B0 is good at correctly detecting dogs' emotions, which could be useful in veterinary care, animal behavior research, and studies of how people interact with animals.