Artificial Intelligence-based Virtual Crime Scene Reconstruction

Abstract

Crime scene investigation is a vital component of forensic science, assisting in uncovering facts behind criminal activities. Traditional crime scene analysis involves manual procedures like photography, sketching, and note-taking, which can be time-consuming, labour-intensive, and susceptible to human errors or oversights. To overcome these challenges, this paper introduces an AI-based virtual crime scene reconstruction system that leverages advanced Machine Learning (ML) and Deep Learning (DL) techniques to analyze visual inputs from crime scenes. The system incorporates two separate yet complementary approaches using powerful vision-language models: BLIP (Bootstrapped Language Image Pretraining) and Google Gemini. These models process crime scene images to identify critical elements (e.g., weapons, bloodstains) and generate detailed, human-readable forensic narratives.