Cat Photo-to-Sticker Translation with Deep Learning (Computer Vision, Generative Modeling)

In this project, we aim to convert real cat images into watercolor-style cat stickers by utilizing image segmentation and image style transfer techniques. We used Mask RCNN for the image segmentation task and used several different models, including neural style transfer, UNIT, CartoonGAN, CycleGAN and StyleGAN2, for the style transfer task. We performed both qualitative and quantitative evaluations for these models.