We investigate the problem of automatically determining what type of shoe left an impression found at a crime scene. This recognition problem is made difficult by the variability in types of crime scene evidence (ranging from traces of dust or oil on hard surfaces to impressions made in soil) and the lack of comprehensive databases of shoe outsole tread patterns. We find that mid-level features extracted by pre-trained convolutional neural nets are surprisingly effective descriptors for this specialized domains. However, the choice of similarity measure for matching exemplars to a query image is essential to good performance. For matching multi-channel deep features, we propose the use of multi-channel normalized cross-correlation and analyze its effectiveness. Our proposed metric significantly improves performance in matching crime scene shoeprints to laboratory test impressions. We also show its effectiveness in other cross-domain image retrieval problems: matching facade images to segmentation labels and aerial photos to map images. Finally, we introduce a discriminatively trained variant and fine-tune our system through our proposed metric, obtaining state-of-the-art performance.
Cross-Domain Image Matching with Deep Feature Maps

Journal: International Journal of Computer Vision
Published: 2019
Primary Author: Bailey Kong
Secondary Authors: James Supancic, Deva Ramana, Charless Fowlkes
Type: Publication
Research Area: Footwear
Related Resources
A New Algorithm for Source Identification of Look-alike Footwear Impressions Based on Automatic Alignment
Presentation at the International Association for Identification
Center for Statistics and Application in Forensic Evidence Update
The information below highlights a sample of current research initiatives led by the CSAFE team. Additional accomplishments in other forensic science disciplines will be discussed in subsequent issues of Forensic…
Automatic Class Characteristic Recognition in Shoe Tread Images
One of the fundamental problems in footwear forensics is that the distribution of class characteristics in the local population is not currently knowable. Surveillance devices for gathering this data are…
Modeling And iNventory of Tread Impression System (MANTIS): The development, deployment and application of an active footwear data collection system
This CSAFE webinar was held on March 24, 2022. Presenters: Dr. Richard Stone Iowa State University Dr. Susan Vanderplas University of Nebraska, Lincoln Presentation Description: This webinar details the development,…