A Convolutional Neural Network (CNN) constructed with PyTorch Lightning is used in this study to provide a new method for identifying emotions. The algorithm categorizes human emotions based on facial image data using deep learning techniques. Ahegao, Angry, Happy, Neutral, Sad, and Surprise are the six emotions that are represented in the dataset. Performance is enhanced through careful training–validation partitioning, data augmentation, and preprocessing. CNN is trained and evaluated using key performance metrics like accuracy and loss, and it has an impressive test accuracy of about 97.85%. These findings show how well the model categorizes emotions and provide useful details about its benefits while highlighting areas that need further work. This study advances the field's body of knowledge and demonstrates how emotion recognition can be enhanced through deep learning.