Comparative Study of Forensic Face Recognition and Fingerprint during Crime Scene investigation and the role of Artificial Intelligence tools in Forensics

Abstract

Surprisingly, the high accuracies previously reported, exceeding 95%, dropped significantly when faced with the more demanding conditions of the forensic scenario, plummeting to as low as 65%. In essence, while facial recognition systems have shown impressive performance in ideal conditions, our study indicates a substantial decrease in accuracy when faced with the complexities and challenges typical of real-world forensic scenarios, highlighting the need for further advancements to bridge this gap. Recent advancements in machine learning and computer vision have shown facial recognition systems achieving accuracies that surpass human performance in controlled settings but fingerprint analysis is proved more accurate in all aspects. To investigate this, we created a large-scale synthetic facial dataset and designed a controlled facial lineup that mimics conditions encountered in real forensic situations. This approach allowed us to systematically assess facial recognition under various challenging real-world conditions. Using both our synthetic dataset and a well-known dataset of actual faces, we tested the accuracy of two widely used neural-based facial recognition systems. Comparative and Analytical method is applied for present Research. Artificial intelligence could help humans in accuracy and speeding up the process of investigation.