Shoe Size Recommendation System Based on Shoe Inner Dimension Measurement Damir OMRČEN, Aleš JURCA UCS, Universal Customization System, Slovenia Abstract The purpose of this work is to develop a system for supporting Internet footwear sales. The goal is to increase the accuracy of the shoe size and model selection when buying in the Internet. We propose a recommendation system which improves the classification accuracy and is based on foot and/or shoe scanners. 271 adult subjects participated in this study. The subjects were asked to estimate the comfort and fit of the shoes. The analysis show that only 50% of selections were correct if the selection relies on subject’s usual size. One of the main reasons for the poor results of shoe selection is the inconsistency of shoe production. Therefore, we developed special devices for measuring the actual inner 3D model of shoes. The shoe inner dimension measurement (SID) is based on x-ray images of the shoes. Based on shoe and foot dimension we are able to increase the recommendation accuracy from 50% up to 85% in some cases. Such a progress could make the Internet shoe sales problematic feasible. Keywords: shoe inner dimension measurement, shoe size recommendation system, fit of footwear, Internet shoe sales 1. Introduction Internet is a multi-billion dollar economy and the Internet's economic potential is growing every year (BBC news, 2010). On the other hand the Internet sales in footwear lag behind. In our opinion the main reason for the poorly developed Internet footwear sales is the inability to estimate the fit of shoes. There have been a number of attempts to make the footwear Internet sales more efficient and successful. In some cases a customer is encouraged to buy more shoe sizes, selects one and returns the others. Sometimes a special electronic or mechanical device is used to measure the feet (Boër, 2007). A good shoe size recommendation system should consider the following components in the recommendation: - geometrical shape of the foot, - personal preferences of the customer, - inner dimensions of the shoe, - technical characteristics (material, construction, ..) of the shoe. Based on the findings we developed UCS recommendation system. One of the crucial components in our recommendation system is a foot scanner. There are number of foot scanners available on the market (Boër, 2007). Some of them use set of cameras to define the exact 3D shape of the foot. The others only produce scan of the sole. In this work we shortly describe a foot scanner, which generates a 2.5 dimensional scan in only a fraction of a second and can be successfully used in a recommendation system. Even more significant component of a good recommendation system is measurement of the inner shape of a shoe. The shoe inner dimension is defined by the shoe last on which the shoe is produced. The shoe last, the most important component of a successful shoe manufacture, has always been the subject of a big interest. Though, a good shoe fitting depends on the finished shoe and not on the last. Namely, the inner dimensions of a shoe might considerably differ from those of the last. There were some minor attempts of measurement of the inner dimensions of shoes. In addition to casting and other destructive methods (which are not useful for a massive implementation) there were some high-tech approaches: CT and optical methods for the inner measurement (Coudert, 2007 and Richter, 2010). Both have significant drawbacks which pushed us to develop a new method (Omrcen, 2010) which is nondestructive, precise enough, fast and economically acceptable also for a massive implementation. In the following we shortly present the basic elements of our development: Non-invasive X-ray based determination of the inner dimensions of a shoe. The main contribution of this study is in development of a recommendation system for supporting Internet footwear sales. The goal is to increase the accuracy of the shoe size and model selection when buying in the Internet. We propose a recommendation system which improves the recommendation accuracy and is based on foot and/or shoe scanners. 2nd International Conference on 3D Body Scanning Technologies, Lugano, Switzerland, 25-26 October 2011 158