Facial emotion recognition is a crucial component of human computer interaction, creating impact on various industries including healthcare, education, and entertainment. The project goal is to develop a robust facial emotion identification system by training a convolutional neural network (CNN) on a dataset of 19,502 images. The integration of PyTorch and TensorFlow helped rapid model development, while the Adam optimizer ensured optimal convergence during training. The capabilities of the system were enhanced by implementing the Haar-cascade algorithm, resulting in a 30% reduction in processing time compared to previous methods. After 100 epochs, each with 98 samples, the proposed methodology achieved an accuracy of 84%, demonstrating a 15% improvement over baseline models. However, the system showed limitations in low-light conditions and with partially occluded faces, where accuracy dropped. Additionally, the model’s performance varied across different emotions, with happiness and surprise being recognized with over 90% accuracy, while more subtle emotions like contempt were sad ness with only 75% accuracy. Future work will focus on addressing these limitations through data augmentation and the incorporation of attention mechanisms to improve performance in challenging scenarios.