Food Image Generation and Translation and Its Application to Augmented Reality

Keiji Yanai, Daichi Horita, Jaehyeong Cho

Abstract: Food image recognition has drawn attention because it is important technology to record and manage people's eating habits automatically. In addition, recently, due to the great progress of generative adversarial networks (GAN) and GAN-based image translation, we can generate food images and change the categories of the foods in food images. In addition, food-to-food image translation can be integrated with augmented reality for new eating experiences as well as food entertainments. In this paper, we introduce some works of ours on food image generation and translation and its application to augmented reality (AR). (1) Conditional food image generation. (2) Food image translation system, "MagicalRice-Bowl" and AR-based system for new eating experience, "DeepTaste". (3) Interactive sketch-based ramen image generating system, "RamenAsYouLike" and its extension with a ramen style encoder.