A Survey of 3D Reconstruction with Event Cameras

TL;DR

This survey delivers the first focused review of 3D reconstruction using event cameras, organizing methods by stereo, monocular, and multimodal input and by reconstruction paradigm (geometry-based, learning-based, and neural rendering with NeRF and 3DGS). It traces the field’s trajectory from traditional geometric techniques to advanced neural rendering, and compiles major datasets and evaluation metrics to support rigorous benchmarking. The authors identify critical gaps in standardization, event representation design, and dynamic-scene real-time reconstruction, and they outline pathways toward scalable, robust, and broadly applicable event-driven 3D reconstruction. The work aims to catalyze progress across robotics, autonomous systems, and immersive media by providing a clear roadmap and a consolidated reference for researchers and developers.

Abstract

Event cameras are rapidly emerging as powerful vision sensors for 3D reconstruction, uniquely capable of asynchronously capturing per-pixel brightness changes. Compared to traditional frame-based cameras, event cameras produce sparse yet temporally dense data streams, enabling robust and accurate 3D reconstruction even under challenging conditions such as high-speed motion, low illumination, and extreme dynamic range scenarios. These capabilities offer substantial promise for transformative applications across various fields, including autonomous driving, robotics, aerial navigation, and immersive virtual reality. In this survey, we present the first comprehensive review exclusively dedicated to event-based 3D reconstruction. Existing approaches are systematically categorised based on input modality into stereo, monocular, and multimodal systems, and further classified according to reconstruction methodologies, including geometry-based techniques, deep learning approaches, and neural rendering techniques such as Neural Radiance Fields (NeRF) and 3D Gaussian Splatting (3DGS). Within each category, methods are chronologically organised to highlight the evolution of key concepts and advancements. Furthermore, we provide a detailed summary of publicly available datasets specifically suited to event-based reconstruction tasks. Finally, we discuss significant open challenges in dataset availability, standardised evaluation, effective representation, and dynamic scene reconstruction, outlining insightful directions for future research. This survey aims to serve as an essential reference and provides a clear and motivating roadmap toward advancing the state of the art in event-driven 3D reconstruction.

Figures (16)

• Figure 1: Roadmap of 3D reconstruction with event cameras. It shows the development from event-based geometry to neural 3D rendering. With advances in technology, event-camera-based 3D reconstruction methods are achieving progressively higher accuracy and realism, enabling more complete and faithful 3D scene rendering.
• Figure 2: Publication trends by category (May 2025)
• Figure 3: Publication timeline in categories (from 2015 to May 2025). The research focus has shifted from early geometric and traditional learning methods to NeRF and 3D Gaussian Splatting, particularly in monocular and multimodal settings, where these emerging techniques have become dominant since 2023.
• Figure 4: Categorisation of methods in the survey. This survey distinguishes stero, monocular, and multimodal event camera systems by their inputs, and further categorises methods by processing type and output type into geometric approaches, learning-based methods, and neural rendering frameworks based on NeRF and 3D Gaussian Splatting.
• Figure 5: Event Camera vs. Traditional Camera. Traditional cameras output images at a fixed frame rate, whereas event cameras respond asynchronously to brightness changes in the scene, continuously generating a stream of events carrying spatial, temporal, and polarity information.