SSNeRF: Sparse View Semi-supervised Neural Radiance Fields with Augmentation


Xiao Cao (National University of Singapore), Beibei Lin (National University of Singapore), Bo Wang (University of Mississippi), Zhiyong Huang (National University of Singapore), Robby T. Tan (National University of Singapore)
The 35th British Machine Vision Conference
PDFPosterVideo (Right click to download)Supplementary

Abstract

Sparse-view NeRF is challenging because limited input images lead to an under-constrained optimization problem for volume rendering. Existing methods address this issue by relying on supplementary information, such as depth maps. However, generating this supplementary information accurately remains problematic and often leads to NeRF producing images with undesired artifacts. To address artifacts and enhance robustness, we propose SSNeRF, a semi-supervised sparse-view NeRF framework based on a teacher-student paradigm. In this framework, the student NeRF is challenged by severe sparse-view degradation while being guided by rectified high-confidence pseudo-labels from the teacher NeRF. Our sparse-view-specific degradations include injecting designed noise into volume rendering weights, perturbing vulnerable layers, and simulating sparse-view blurriness. These techniques together force the student NeRF to recognize degradation caused by sparse-view training data, ultimately helping the model effectively handle the noise and incomplete information inherent in sparse views. Extensive experiments demonstrate the effectiveness of our SSNeRF in generating novel views with less sparse-view degradation. We will release code upon acceptance.

Citation

@inproceedings{Cao_2025_BMVC,
author    = {Xiao Cao and Beibei Lin and Bo Wang and Zhiyong Huang and Robby T. Tan},
title     = {SSNeRF: Sparse View Semi-supervised Neural Radiance Fields with Augmentation},
booktitle = {36th British Machine Vision Conference 2025, {BMVC} 2025, Sheffield, UK, November 24-27, 2025},
publisher = {BMVA},
year      = {2025},
url       = {https://bmva-archive.org.uk/bmvc/2025/assets/papers/Paper_81/paper.pdf}
}


Copyright © 2025 The British Machine Vision Association and Society for Pattern Recognition
The British Machine Vision Conference is organised by The British Machine Vision Association and Society for Pattern Recognition. The Association is a Company limited by guarantee, No.2543446, and a non-profit-making body, registered in England and Wales as Charity No.1002307 (Registered Office: Dept. of Computer Science, Durham University, South Road, Durham, DH1 3LE, UK).

Imprint | Data Protection