Deep Learning and Computer Vision for Forensic Analysis and the Interpretation of Footwear Marks.

Footwear marks – impressions left by the sole of a shoe – are a key evidence type used to associate footwear to a place where a criminal activity may have occurred or as an intelligence tool used to link places together. In the forensic examination of footwear marks, the sole patterns examined can often be distorted or partial and the linkage of these to a shoe can be challenging. Currently subjective (by eye) observations are used in such examinations.  

We wish to examine the potential of computer vision and machine learning (including deep learning) to link footwear marks to the shoes that made them. This will involve the development of algorithms to provide probabilistic matches, and their validation on datasets acquired specifically for this purpose.  

The research is funded by the Leverhulme Trust and will be based at the Leverhulme Research Centre for Forensic Science (Dr R Such-Solis and Prof N Nic Daeid) in collaboration with the Computer Vision and Image Processing (CVIP) group in Computer Sciences & Informatics (Prof S McKenna and Prof E Trucco) and will build on a rich platform of experience and software for image analysis, segmentation, and classification. The researcher will also work with forensic science practitioners in the relevant fields.