A common goal of forensic shoeprint analysis is to identify shoe models or designs that are similar to that of a given print, such as a print found at a crime scene. Quantifying similarity between shoe outsole patterns is difficult because it requires both a set of well-defined features and an accurate method to classify outsoles according to those features. A set of geometric features was developed based on common geometric shapes, such as circles and quadrilaterals. A new classifier was then trained for the pretrained convolutional neural network base of VGG16 to create a model, named CoNNOR (Convolutional Neural Network for Outsole Recognition), to automatically classify portions of outsole images into the new geometric scheme of outsole class characteristics. During the analysis of CoNNOR’s performance, new diagnostic plots were developed which provide a better method to assess classification errors of multi-class, multi-label models. In general, CoNNOR performs well on images with unambiguous shapes and moderate color contrast; additional improvements may be realized by preprocessing the images to improve contrast as well as by integrating spatial relationships between geometric features. CoNNOR represents a significant improvement to the current manual classification of footwear tread patterns, facilitating new modes of data collection and automatic processing that will expand the data available for assessment of footwear class characteristic frequency in the population.
A convolutional neural network for outsole recognition
Journal: Creative Component
Published: 2019
Primary Author: Miranda Tilton
Secondary Authors: Susan Vanderplas
Type: Publication
Research Area: Footwear
Related Resources
Graph-Theoretic Techniques for Forensic Image Comparisons
This presentation is from the 76th Annual Conference of the American Academy of Forensic Sciences (AAFS), Denver, Colorado, February 19-24, 2024.
ShoeCase: A data set of mock crime scene footwear impressions
This project’s main objective is to create an open-source database containing a sizeable number of high-quality images of shoe impressions. The Center for Statistics and Applications in Forensic Evidence (CSAFE)…
A finely tuned deep transfer learning algorithm to compare outsole images
In forensic practice, evaluating shoeprint evidence is challenging because the differences between images of two different outsoles can be subtle. In this paper, we propose a deep transfer learning-based matching…
An automated alignment algorithm for identification of the source of footwear impressions with common class characteristics
We introduce an algorithmic approach designed to compare similar shoeprint images, with automated alignment. Our method employs the Iterative Closest Points (ICP) algorithm to attain optimal alignment, further enhancing precision…