Deep learning for recipe generation

Abstract

Food, as we all know, is a very important part of a culture and recipes are a major source of useful information in our daily lives. From conventional cookbooks to YouTube cooking videos, one can find numerous cooking ideas and practical instructions readily available through various channels. Finding or recommending recipes that meet our needs is, however, not always an easy task. Researchers have studied many intelligent ways of meeting such needs using computational methods. Deep learning has recently opened up interesting and powerful ways to exploit the vast troves of recipes and food images amassed in digital forms and support innovative use cases in food- and health-related applications. In his chapter on ‘Deep Learning for Recipe Generation’, Professor Chan Sin-wai of Caritas Institute of Higher Education provides an overview of these use cases, reviewing the use of artificial intelligence technology in the visualization of Chinese recipes with English translations through text (recipe)-to-image and image-to-text (recipe) generators and make human and automatic evaluations of their outputs and quality.