AI-Emotion Image Generator by using a CNN Model

Artificial Intelligence (AI) is increasingly moving beyond analytical tasks toward understanding and expressing human emotions. This study presents an AI-based Emotion Image Generator that unites the strengths of Convolutional Neural Networks (CNNs) for emotion detection and Generative Adversarial Networks (GANs) for image creation. The system first identifies facial expressions and categorises them into seven basic emotions— happiness, sadness, anger, surprise, disgust, fear, and neutrality. Once the emotion is recognised, it is passed as a conditioning parameter to the GAN, producing a corresponding expressive or artistic image. The model was trained on well-known datasets such as FER2013 and AffectNet, ensuring broad coverage of real-world facial variations. Experiments show that the CNN model achieved an emotion-recognition accuracy of 93.7%. At the same time, the GAN produced realistic emotion-specific images with a Fréchet Inception Distance (FID) of 24.6 and a Structural Similarity Index (SSIM) of 0.87. The results confirm that combining CNN and GAN architectures can effectively bridge emotional understanding and creative image synthesis, making it valuable for interactive media, virtual avatars, and affective computing applications.