The quick advancement of deep learning methods and artificial intelligence has made it feasible to identify emotions and feelings based on linguistic and visual data. Human-computer interaction systems are better able to comprehend human emotions and intentions and react appropriately thanks to these detection and recognition techniques. On the other side, deep learning models are also used to categorize pets facial expressions, and pet owners must pay attention to their pets' emotions in order to modify and control their behavior in a timely manner. In this study, we conduct an experiment to predict dogs emotional states, which are divided into three categories happy, relaxed, and angry. This research combine the Convolution Neural Network and  Vision Transformer model to assess and categorize canine emotions known as   Compact Convolution Transformer .This achieved the highest accuracy and promising approach for efficient and accurate emotion recognition in dogs as compare to Convolution Neural Network Models and Vision Transformer Models.

Keywords—ConvolutionTransformer,Linear Embedding, Data Augmentation,Vission transformer,Encoder