VoxNeRF: Bridging Voxel Representation and Neural Radiance Fields for Enhanced Indoor View Synthesis
Sen Wang Qing Cheng Stefano Gasperini Wei Zhang Shun-Cheng Wu Niclas Zeller Daniel Cremers Nassir Navab
Technical University of Munich Huawei Karlsruhe University of Applied Sciences
VoxNeRF improves neural indoor reconstruction and novel view synthesis by combining voxel-encoded geometry priors with neural radiance fields. An efficient voxel-guided sampling strategy focuses computation on the most relevant ray segments, and a robust depth loss improves quality in sparse-view settings. Experiments on ScanNet and ScanNet++ show strong results for indoor interpolation and extrapolation.
Paper
@ARTICLE{Wang2025VoxNeRF,
author={Wang, Sen and Cheng, Qing and Gasperini, Stefano and Zhang, Wei and Wu, Shun-Cheng and Zeller, Niclas and Cremers, Daniel and Navab, Nassir},
journal={IEEE Robotics and Automation Letters},
title={VoxNeRF: Bridging Voxel Representation and Neural Radiance Fields for Enhanced Indoor View Synthesis},
year={2025},
volume={10},
number={6},
pages={5903-5910},
doi={10.1109/LRA.2025.3559844}
}