Food plating referencer: A Visual Plateware Selection System Based on Image Masking and Similarity Search

Visually appealing food presentation enriches the dining experience. The impression a dish (in this paper ``dish'' is used exclusively to denote plated food) creates is dramatically influenced by the food, its plating, the tableware, and other furnishings. This research aims to realize visually enjoyable dining by diversely enhancing dishes through various plating methods on a selection of crockery. The proposed system takes user-owned or store-sold plates as ``plates accessible to users'' queries and displays food images utilizing similar plates. Conversely, specific foods can be used as queries to display images of dishes used the plate similar to those accessible to users. From this, users would gain inspiration and plating suggestions for combining menu items with plates accessible to them. The system includes pairs of ``dish images'' and ``images with only the extracted plate portions of those dishes'' as its dataset. Based on color information, the query plate image is used for a similarity search among the plate images in the dataset. This identifies images of dishes with plates similar to the query plate. The dataset’s images have the plates’ centers removed, whereas the query plate images do not. We conducted preliminary experiments on whether this difference affects similarity calculations. Based on those results, this paper verified whether computational similarity evaluation matches human similarity evaluation as a first step toward measuring the system’s usefulness.