Rendered results from our dataset/simulator.

Abstract

We present 3D Gaussian Splatting to Event Generation (GS2E), a large-scale event synthetic dataset designed for high-fidelity event vision tasks from real-world sparse multi-view RGB images. Existing event datasets either synthesize from dense RGB videos, which lack viewpoint diversity and geometric consistency, or rely on expensive hardware setups that are difficult to scale. GS2E addresses these limitations by leveraging 3D Gaussian Splatting to reconstruct photorealistic static scenes and by introducing a physically-informed simulation pipeline that integrates adaptive trajectory interpolation and physically-consistent event contrast threshold modeling. This approach produces temporally dense and geometrically consistent event streams across diverse motion and lighting conditions, while preserving strong alignment with underlying scene structures. Experimental results on event-based 3D reconstruction and deblurring tasks demonstrate the improved generalization and practical value of GS2E as a benchmark for event vision.

Citation

TBD
        

Acknowledgements

Credit template source: Michaël Gharbi. Thanks to your team, institution, and any collaborators.