Human communication relies on emotions, either verbally or non-verbally through cues like hand gestures, facial expressions, etc. These non-verbal cues significantly contribute to enhancing the understanding of the communicated message. Facial expressions convey emotions, augmenting the comprehension of conversations. Moreover, these facial cues not only enhance human-to-human interactions but also human–computer interactions, providing deeper insights into human emotions. Given the advancements in AI, there is a growing demand for systems capable of recognizing facial emotions. In this study, effective facial emotion recognition is performed using convolutional neural networks (CNNs). In this work, facial emotions are classified using two different CNN models. Of these two models, the first one comprises of four convolution layers followed by four max pooling layers, and two fully connected layers. Images undergo pre-processing, involving conversion to grayscale and normalization, followed by face detection using the Haar cascade algorithm. The detected facial images are then input into the CNNs for emotion classification, categorizing emotions into seven categories: anger, neutral, disgust, fear, happy, sadness, and surprise. The FER 2013 dataset is employed for training and evaluation. Results indicate that the four-layer model achieves a 74% accuracy rate, while the eight-layer model achieves an 84% accuracy rate.