Anime-Style Image Translation Using GANs

Practicing - Pytorch / GAN in Python

This project explores the application of Generative Adversarial Net- works (GANs) for the task of transforming real-world images into anime- style images. We implemented a GAN architecture that leverages Residual Blocks within the generator to enhance the translation. The dataset we choose comprises real-world images from COCO dataset and corresponding anime-style from https://www.kaggle.com/alamson/safebooru, with the latter undergoing a smoothing process for improved style consistency. The results demonstrate the model’s ability to translate the style of input images, presenting a novel application of GANs in the domain of animate image transformation.