Enhancing the Classification Accuracy using CNN, GANs, and data Augmentation Techniques

This research paper’s primary objective revolves around addressing the critical issue of handling imbalanced data sets, particularly in the context of image classification tasks with an uneven distribution of images. The imbalanced data set can lead to biased predictions, which can have adverse impacts in numerous domains such as medical diagnostics and agriculture. This clearly shows the urgency of addressing this problem and finding effective solutions to mitigate its impact. Hence, this study began with an imbalanced image data set, resulting in biased accuracy, and to address this problem, this research has proposed an innovative two-step approach using data augmentation and generative adversarial networks (GANs) to balance the data set. This balanced data set was then used to train the CNN classification model, achieving impressive results with an accuracy of 97.03%, a precision of 0.75, a recall of 0.83, and an F1-score of 0.79. Hence, the proposed methodology significantly improves classification accuracy and performance, demonstrating the potential of this approach to rectify imbalances in image classification tasks.