Enhancing forensic shoeprint analysis: Application of the Shoe-MS algorithm to challenging evidence.

Area of Science:

• Forensic Science
• Computer Science
• Artificial Intelligence

Background:

• Quantitative assessment of pattern evidence is crucial in forensic investigations.
• Deep learning offers promising tools for pattern recognition and analysis.
• Forensic footwear analysis requires accurate source identification and classification.

Purpose of the Study:

• To explore the Shoe-MS algorithm, a deep learning framework for forensic footwear analysis.
• To evaluate the algorithm's performance in source identification and classification of degraded images.
• To assess Shoe-MS's utility in aiding forensic examiners.

Main Methods:

• The study implemented the Shoe-MS algorithm, a deep learning framework.
• Input: two paired images of footwear.
• Output: a similarity score between 0 and 1.

Main Results:

• Shoe-MS demonstrated high performance in both source identification and degraded image classification tasks.
• The algorithm produced reliable similarity scores, supporting probabilistic and repeatable assessments.
• High accuracy was observed even with low-quality crime scene images.

Conclusions:

• Shoe-MS is a valuable tool for forensic examiners evaluating pattern evidence.
• The algorithm enhances the reliability and reproducibility of forensic footwear analysis.
• Shoe-MS shows significant potential, particularly for analyzing degraded or incomplete evidence.