A Deep Learning Approach for Improving Acne Type Classification Accuracy Using Hybrid Transfer Learning and Synthetic GAN-Based Augmentation

Acne vulgaris, a common dermatological disorder that affects adolescents and young adults, often has social and psychological impacts. Accurate classification of acne types is crucial for efficient diagnosis and treatment, but traditional approaches face difficulties due to the complex shape of lesions, skin color differences, and uneven lighting. This study proposes a hybrid deep learning system that combines Deep Convolutional Generative Adversarial Network (DCGAN) for synthetic data augmentation with Transfer Learning using pretrained CNN architectures, namely VGG16 and ResNet50. The approach addresses dataset imbalance across six acne types: Blackheads, Whiteheads, Papules, Pustules, Cyst, and Comedo, generating realistic synthetic images to improve training diversity and reduce overfitting. Experiments demonstrate substantial performance improvements after DCGAN augmentation, with VGG16 achieving 84.7 % accuracy and ResNet50 48.2 %, significantly higher than baseline models trained without augmentation. Confusion matrix analysis shows enhanced classification stability, particularly for minority classes such as Whiteheads and Cyst. These findings show that incorporating generative augmentation and transfer learning successfully enhances model robustness and generalization in situations with limited data. This developed approach efficiently supports dermatologists in making evidence-based clinical decisions by providing a reliable and scalable automated acne classification solution. This hybrid method lays the foundation for future developments that leverage Explainable Artificial Intelligence (XAI) and advanced generative models to improve readability and applicability in smart dermatology systems, particularly in remote or resource-limited healthcare settings. This study demonstrates that AI-assisted diagnostic tools can significantly improve dermatological care by reducing clinician workload and fostering more consistent, reliable, and equitable patient outcomes.