Application of Generative Adversarial Networks in Brahmi Script Recognition

Abstract

This chapter examines the application of Generative Adversarial Networks (GANs) for the recognition and preservation of the Brahmi script, an essential component of cultural and historical heritage. It covers the script's significance, the challenges of ancient script recognition, and introduces GANs, including DCGAN, CycleGAN, and StyleGAN, highlighting their strengths in generating synthetic data, improving image quality, and pattern recognition. The chapter details the methodology of data collection, GAN architecture modifications, and training processes, emphasizing performance metrics like Structural Similarity Index (SSIM), Peak Signal-to-Noise Ratio (PSNR), accuracy, precision, recall, and F1-score. The findings show StyleGANs excel in image quality while DCGANs perform better in accuracy and precision. Practical applications discussed include enhancing historical research, creating digital archives, and developing educational tools. The chapter concludes with future research directions for further improvements in script recognition and preservation.