Criminator: An Easy-to-Use XR "Crime Animator" for Rapid Reconstruction and Analysis of Dynamic Crime Scenes

Paper Title

Criminator: An Easy-to-Use XR 'Crime Animator' for Rapid Reconstruction and Analysis of Dynamic Crime Scenes

Publication Info

• Topic area: XR-based tools for crime scene reconstruction and analysis.
• Keywords: XR, crime scene reconstruction, forensic animation, virtual reality, usability, co-design, forensic investigation, hypothesis testing, courtroom presentation, police training.

Background and Problem

• Problem / challenge: Current tools for crime scene reconstruction focus on static 3D models and are inaccessible to non-experts due to complexity and cost. Temporal relationships and dynamic event reconstruction remain underexplored, limiting usability for hypothesis testing and courtroom presentations.
• Significance: Accurate and accessible crime scene reconstruction tools could democratize forensic analysis, improve judicial processes, and enhance training for law enforcement, while reducing contamination risks and costs associated with on-site investigations.
• Motivation and related work: Prior research has explored XR for static crime scene reconstruction and training but lacks tools for dynamic event reconstruction. Existing animation tools are complex, requiring advanced skills. This paper addresses the gap by developing a user-friendly XR-based crime scene animator.

Solution

• Proposed approach: Criminator—a methodological framework and XR toolkit for rapid prototyping and animation of dynamic crime scenes, designed to be accessible to non-experts.
• Novelty:
  1. Co-design process with forensic experts to identify requirements and refine features.
  2. Development of a flexible animation authoring framework inspired by video editing tools, enabling intuitive creation of dynamic crime scenes.
  3. Integration of Gaussian Splatting for high-fidelity 3D environment rendering using phone-based scanning.
  4. Evaluation of usability and effectiveness with both trained criminologists and lay participants.
• Procedure and key techniques:
  • Co-design process with forensic experts using iterative prototyping (low-fi sketches, Wizard-of-Oz prototypes, high-fidelity prototypes).
  • Implementation of an animation editor with track-based mechanisms for managing effects and props.
  • Use of controllers for embodied animation creation and navigation in VR.
  • Integration of effects like full-body tracking, rigid transform, and interactive transform for dynamic scene reconstruction.

Results

• Concrete findings:
  • Stage 1 usability (SUS): Mean score of 76.53 (B grade), indicating above-average usability.
  • Stage 2 usability (SUS): Mean score of 66.67 (C grade), reflecting moderate usability for animation authoring.
  • Animation quality ratings: Trained participants performed slightly better, but untrained participants also created acceptable animations.
• Advantage over baselines: Simplifies animation creation compared to desktop software, enabling non-experts to replicate dynamic crime scenes with minimal training.
• Experiments / evaluation:
  • Two-stage user study with 18 participants (6 trained criminologists, 12 laypeople).
  • Tasks included observing pre-generated animations and authoring new animations in VR.
  • Metrics: SUS, NASA-TLX, qualitative feedback, animation quality ratings, gaze and location patterns.
• Limitations and future work:
  • Small sample size of trained experts due to recruitment challenges.
  • Limited accuracy in lower-body animations due to SDK constraints.
  • Need for further validation of legal acceptability and uncertainty visualization.
  • Future work to include judges, lawyers, and policymakers in design and evaluation.

Summary

Criminator is an XR-based toolkit designed to democratize dynamic crime scene reconstruction by enabling non-experts to create animations intuitively. Evaluated through a user study, the tool demonstrated usability for observation and animation tasks, with potential applications in hypothesis testing, courtroom presentations, and police training. While usability challenges and legal concerns remain, the framework shows promise for enhancing forensic analysis and judicial processes. Future research should expand stakeholder involvement and explore advancements in XR and computer vision technologies.