Mango Generative Adversarial Network

This research aims to achieve the objectives of formulating a deep generative adversarial network (GAN) for translating sketches to "Sala" mango images and vice versa, developing a web application for sketch-to-image transformation, and evaluating the quality of the generated images. The methodology involves data collection, preprocessing, and augmentation, followed by the implementation of an image-to-image conditional GAN. The Frechet Inception Distance (FID) metric is utilized to assess image quality. The findings demonstrate the successful translation of sketches to high- quality "Sala" mango images, with the web application providing a user-friendly interface for convenient transformation. The low FID scores indicate the close resemblance of the generated images to the ground truth, highlighting the effectiveness of the GAN model. This research contributes to the advancement of image generation techniques and provides a valuable tool for transforming sketches into realistic mango images, enhancing the field of generative adversarial networks and mango imagery.