VeriScene reconstructs crime scenes from photos and statements
VeriScene: Reconstructing Crime Scenes from Legal Evidence via World-Model Agent
Computer Vision and Pattern RecognitionCryptography and Security
Summary
Piecing together how a crime happened can be tricky because evidence comes in many forms like pictures and witness stories, which can sometimes conflict or be unreliable. The authors created VeriScene, a tool that carefully combines this kind of legal evidence into a believable 3D reconstruction that follows the rules of physics. VeriScene checks and improves its ideas by testing different scenarios, making sure every part matches the evidence and creates a plausible video reenactment of the crime. Their tests show VeriScene is better than previous approaches at including more evidence and sticking to the facts.
What this means in practice
- •For forensic investigators: Reconstruct complex crime scenes by integrating photos and witness accounts into a consistent physical scenario for case analysis.
- •For legal teams: Create video reenactments of crimes that cite all evidence and verify physical possibilities to support courtroom presentations.
Authors
Kevin Chuanpu Fu, Yongsen Zheng, Zee Kin Yeong, Kwok-Yan Lam
Abstract
World models take multimodal inputs like text, photos, and diagrams to generate dynamic scenes in accordance with the laws of physics, thus opening a compelling application: fusing multimodal legal evidence to re-create a crime scene and re-enact how an offence could have been committed. However, feeding the raw, unorganized evidence into a world model fails in forensic use: it silently drops evidence, glosses over contradictory testimony, and produces motion that violates the evidentiary record. This paper presents VeriScene, an agent that orchestrates the world model: it reconstructs crime scenes from forensic photographs and witness statements of varying reliability, keeping every claim traceable to evidence and every motion physically plausible. VeriScene iteratively fuses the evidence into a cited narrative under an auditing loop, verifies the hypothesized dynamics via probe rollouts in the world model with corrective constraint injection, and renders the offence as a re-enactment video from a fused keyframe. On a benchmark of 25 crime scenarios across 7 physically-driven case types (139 forensic-style photographs and 65 statements with planted unreliability), VeriScene attains 0.9014 evidence coverage and 0.7217 factual consistency (0-1 scale) on the 20 test scenes, outperforming an end-to-end multimodal-LLM baseline by 20.35% in factual consistency and 34.88% in temporal coherence, while generalizing across four LLM orchestration backends at USD 1.82 per scene.