Modelling wear patterns on footwear outsoles

Abstract

ABSTRACT: The outsoles of footwear develop nicks, cuts, and tears via repeated exposure to the abrasive forces that occur between the outsole and the ground. These abrasions result in the formation of characteristics unique to the outsole and the individual wearing them; additionally resulting in the degradation of the outsole design imprinted by the manufacturer. The combination of these characteristics allow the forensic scientist to uniquely identify the individual to whom it belongs. Quite often a period of time can elapse between the discovery of a shoeprint at the crime scene and the identification of a suspect. In these instances, the forensic scientist must rely on their training and expertise—developed through years of experience and study—to determine if the crime scene shoeprint matches the out-sole of the suspect’s shoe.

This work introduces a computational framework capable of modelling wear patterns on the out-soles of footwear. This model is able to predict the evolution of the wear pattern after an arbitrary time period given in weeks. We introduce an additional model capable of reconstructing the outsole back to its original state on a given week. This framework—built on convolutional neural networks—provides an objective point of reference for forensic scientists in their evaluation of outsole wear patterns.