Feasibility of Neural Radiance Fields for Crime Scene Video Reconstruction

TL;DR

The paper addresses the feasibility of applying Neural Radiance Fields (NeRF) variants to reconstruct crime scenes from video data. It surveys three core innovations—Multi-Object Synthesis, Deformable Synthesis, and Lighting enhancements—and evaluates how each handles static vs. dynamic content and illumination variability for novel view synthesis. By discussing related works such as NeRF-MS, NeRF-W, DS-NeRF, NeRD, and MPI-based approaches, the authors assess progress and identify gaps in wild-data robustness, multi-object dynamics, and full-scene relighting. They conclude that full crime-scene reconstruction from video is not yet feasible, yet the analyzed trends suggest improving feasibility with better datasets, dynamic-object modeling, and comprehensive lighting capabilities in the future.

Abstract

This paper aims to review and determine the feasibility of using variations of NeRF models in order to reconstruct crime scenes given input videos of the scene. We focus on three main innovations of NeRF when it comes to reconstructing crime scenes: Multi-object Synthesis, Deformable Synthesis, and Lighting. From there, we analyse its innovation progress against the requirements to be met in order to be able to reconstruct crime scenes with given videos of such scenes.