Approaches to Food/Non-food Image Classification Using Deep Learning in Cookpad

In this paper we report our approach to image classification, in particular to the food/non-food image classification problem, as used by our Cooking Log product of Cookpad Inc. We augment our existing services with an architecture which includes a loosely-connected and asynchronous image analysis module. One challenge of the classification is that the non-food class is very vast and varied and can only be defined in context. We find that having the nonfood class consist of multiple subclasses effectively improves both precision and recall by capturing different types of features in the images of that class.