3D Interactive Lineups (3DIL)
Increasing eyewitness identification accuracy in lineups using 3D interactive virtual reality.
The 3D Interactive Lineups (3DIL) project is dedicated to enhancing eyewitness identification accuracy through cutting-edge 3D virtual reality (VR) technology. Traditional 2D photo lineups often fail to capture the complexity of real-life face recognition, leading to mistaken identifications and potential wrongful convictions. The 3DIL project aims to develop interactive 3D lineups, allowing witnesses to view lifelike facial models from multiple angles, increasing the realism and reliability of identifications. Project Website
The research involves a cross-national team from the UK, Germany, and Canada, combining expertise in psychology, computer science, and law enforcement practices. Our collaborative approach ensures robust testing and practical relevance across different legal systems and cultural contexts.
Publications
In this project, the Humboldt University is responsible for synthesizing and capturing photorealistic 3D stimuli. The resulting publications mainly focus on 3D GANs and 3D Gaussian Splatting.
[VISAPP '25] Improving Adaptive Density Control for 3D Gaussian Splatting
In this work we propose three simple modifications to the 3DGS optimization algorithm to improve the rendering quality. Our method especially improves the visual quality in background regions and in smooth areas where only few Gaussians are needed.
[CVPRW '24] Gaussian Splatting Decoder for 3D-aware Generative Adversarial Networks
In this work, we present a novel approach that combines the high rendering quality of NeRF-based 3D-aware Generative Adversarial Networks with the flexibility and computational advantages of 3DGS. By training a decoder that maps implicit NeRF representations to explicit 3D Gaussian Splatting attributes, we can integrate the representational diversity and quality of 3D GANs into the ecosystem of 3D Gaussian Splatting for the first time.
[VISAPP '24] Multi-view Inversion for 3D-aware Generative Adversarial Networks
Our method builds on existing state-of-the-art 3D GAN inversion techniques to allow for consistent and simultaneous inversion of multiple views of the same subject. We employ a multi-latent extension to handle inconsistencies present in dynamic face videos to re-synthesize consistent 3D representations from the sequence.
Funded by the Deutsche Forschungsgemeinschaft (DFG, German Research Foundation) – 502864329
