Understanding Dog Emotions Through Deep Learning: A CNN-based Classification Framework

Our research article focuses on developing a deep learning-based Convolutional Neural Network (CNN) model for classifying dog emotions into four categories: sadness, happiness, relaxation, and anger. The significance of understanding canine emotions is highlighted, emphasizing the need for an accurate classifier to enhance animal welfare and the human-animal bond and Quality of life. We utilize a dataset collected from Kaggle and implement a CNN architecture, trained using the Stochastic Gradient Descent (SGD) optimizer. The methodology includes detailed descriptions of dataset preprocessing, model creation, and training procedures. Our experimental setup involves evaluation metrics, dataset partitioning, and computational resources. Results demonstrate the CNN model’s superior performance, achieving a validation accuracy of 99.60%, significantly outperforming other baseline models like Support Vector Machine (SVM) and K-Nearest Neighbors (KNN), which achieved accuracies of $89 \%$ and $81 \%$, respectively. Practical applications of our model in veterinary care and human-dog interactions are discussed. Future directions suggest avenues for further research, including expanding the dataset and exploring different model architectures. The conclusion summarizes the key findings and emphasizes the implications for enhancing the understanding and interaction between humans and dogs.