Facial emotion classification is critical in improving human-computer interaction and applications like affective computing and mental health monitoring. Accurate recognition of emotions from images enables systems to better understand and respond to human emotional states. However, existing algorithms face challenges in capturing subtle emotional expressions, handling variations in facial features due to lighting, angles, or occlusion, and generalizing effectively across diverse datasets. EmoNet addresses these challenges through its CNN-based architecture, designed specifically for emotion classification from facial images. It utilizes convolutional layers to extract spatial features, followed by fully connected layers for classifying emotions. EmoNet excels in processing complex facial expressions, adapting to different lighting and occlusion conditions, and generalizing across varied datasets. This makes it highly suitable for real-time applications requiring accurate emotion recognition. The performance of EmoNet is evaluated using metrics such as accuracy, precision, recall, and F1-score, which highlight its effectiveness in handling diverse real-world scenarios. Notably, EmoNet achieves an impressive accuracy of 99.6% and a Kappa score of 94.1, establishing it as the top performer across all emotional categories. © 2025 Elsevier B.V., All rights reserved.