Artificial intelligence powered crime scene analysis service.

Area of Science:

• Forensic Science
• Artificial Intelligence
• Computer Vision

Background:

• Crime remains a significant societal issue, necessitating efficient investigative processes.
• Traditional forensic methods are often manual, time-consuming, and prone to human error.
• There is a need for advanced technological solutions to support law enforcement and expedite justice delivery.

Purpose of the Study:

• To develop and evaluate a novel web application integrating artificial intelligence for streamlining crime scene investigations.
• To automate key forensic tasks including fingerprint reconstruction, weapon detection, and human activity recognition.
• To provide forensic investigators with an efficient tool for analyzing crime scene data.

Main Methods:

• Development of a web application with an intuitive interface for uploading crime scene images and videos.
• Implementation of an autoencoder model for fingerprint reconstruction.
• Utilization of YOLO NAS for object detection (weapon recognition).
• Application of VGG16 for human activity recognition.

Main Results:

• The fingerprint reconstruction model achieved a validation loss of 0.0477 and step loss of 0.0487, improving the analysis of partial fingerprints.
• YOLO NAS demonstrated a mean Average Precision (mAP) of 77.8% for weapon detection.
• VGG16 achieved a high accuracy of 98.21% for human activity recognition.

Conclusions:

• The AI-driven web application significantly enhances the efficiency and accuracy of crime scene investigations.
• Automated analysis of fingerprints, weapons, and human activities aids in identifying perpetrators and understanding events.
• This technology offers a promising advancement for forensic science and the administration of justice.