With only a limited number of footwear experts in existence to help identify the make and model of footwear impressions at crime scenes, a team of researchers from Bournemouth University in the U.K. have developed a neural network for identifying the make and model of footwear impressions.

Developed in conjunction with Bluestar Software, the neural network, which is a type of artificial intelligence (AI), examines black and white footwear impressions, automatically recognizing the shape of component treads, logos or writing included in the shoe’s impression — each of which links to a code in a classification system.

According to the team, the codes are used to search the footwear impression database and the AI offers suggested codes for the user to verify.

During a series of experiments, a so-called occasional user — a non-footwear expert such as forensic and police personnel using the footwear database — was tasked with analyzing 100 randomly selected shoe prints, correctly identifying the footwear impressions between 22% and 83% of the time. Meanwhile, the AI was between 60% and 91% successful at identifying the footwear. However, actual footwear experts were correct roughly 100% of the time.

As such, the researchers suggest that although the AI won’t replace the footwear impression experts, it could be used to help free up actual experts so that they can focus their efforts on more difficult-to-solve cases.

Additionally, another neural network was created and tasked with identifying the gender of footprints. According to its developers, the AI was 90% accurate, outperforming a podiatrist who performed the same task and only successfully identified the gender of footprints 50% of the time.

The research, Sexing white 2D footprints using convolutional neural networks, appears in the journal PLOS ONE.

To contact the author of this article, email mdonlon@globalspec.com