AI Agents - 3D Crime Scene Investigation

This research project is carried out in cooperation with the Bavarian State Criminal Police Office (BLKA) and addresses the generation of a so called “AI-Agent”. The overarching goal is the deep learning-based analysis of 3D topography in both 2D images and 3D data in the context of digital crime scene twins. This includes, in particular, the recognition of specific objects relevant to the crime scene and the data-driven investigation of their global spatial relationships in a geometrically accurate yet photorealistic, physically correct, and complete twin of the crime scene. This ultimately allows detailed forensic analyses and constructive investigations to be carried out by a visually present (avatar), responsive (text-to-scene interaction), and appropriately trained AI agent.

Research areas:

• 3D Reconstruction
• Digital Twins
• (Large-)Scene Interpretation
• Text-to-Scene Interaction
• Anomaly detection
• Physics Simulation and Integration in 3D

Other Projects

Autonomous Landing Spot Detection for UAVs
Research collaboration with Institute of Flight System Dynamics (TUM) - Development of demo software for LiDAR- and RGB-driven autonomous landing site detection for drones