Generative Adversarial Networks for Data Augmentation in Image Recognition

Abstract

This study investigates the application of Generative Adversarial Networks (GANs) for data augmentation to enhance image recognition performance. Synthetic images produced by GANs were incorporated into existing labeled datasets to improve model training, particularly under conditions of limited labeled data availability. The primary objective was to assess whether GAN-generated data could strengthen the model's generalization capabilities. Accuracy served as the principal performance metric for evaluating the proposed approach. Experimental results demonstrate that the integration of realistic synthetic images significantly expands the training dataset, thereby improving recognition accuracy. These findings suggest that GAN-based augmentation is a valuable strategy in scenarios were acquiring large volumes of labeled image data is impractical.