This research basically investigates the development of a pet emotion recognition system utilizing convolutional neural networks (CNNs), specifically examining the Mobile Net, VGG, and Dense Net architectures. This study entails the creation and preprocessing of a diverse dataset containing 2050 images of pets expressing various emotions. Through transfer learning, the chosen CNN architectures are fine-tuned for the specialized task of classifying pet emotions. The models undergo training, assessment, and comparison based on performance metrics such as accuracy, precision, recall, and F1 score. An accuracy of 82% was attained using Dense Net, while Mobile Net V2 achieved an accuracy of 89.50%. VGG-16 and VGG-19 models yielded accuracies of 75.50%, demonstrating the varying performance of different architectures in the context of facial emotion recognition. Notably, Mobile NetV2 emerged as the leading performer, boasting an impressive accuracy rate of 89.50%. The research acknowledges the unique challenges associated with recognizing pet emotions owing to their diverse appearances and expressions. The comparative analysis of Mobile Net, VGG, and Dense Net architectures provides valuable insights for researchers and practitioners aiming to implement effective pet emotion recognition systems.