Fast Sparse View Guided NeRF Update for Object Reconfigurations

Ziqi Lu1, Jianbo Ye2, Xiaohan Fei2, Xiaolong Li2, Jiawei Mo2, Ashwin Swaminathan2, Stefano Soatto2
1MIT 2Amazon AWS AI Labs
Arxiv

Abstract

Neural Radiance Field (NeRF), as an implicit 3D scene representation, lacks inherent ability to accommodate changes made to the initial static scene. If objects are reconfigured, it is difficult to update the NeRF to reflect the new state of the scene without time-consuming data re-capturing and NeRF re-training. To address this limitation, we develop the first update method for NeRFs to physical changes. Our method takes only sparse new images (e.g. 4) of the altered scene as extra inputs and update the pre-trained NeRF in around 1 to 2 minutes.

Particularly, we develop a pipeline to identify scene changes and update the NeRF accordingly. Our core idea is the use of a second helper NeRF to learn the local geometry and appearance changes, which sidesteps the optimization difficulties in direct NeRF fine-tuning. The interpolation power of the helper NeRF is the key to accurately reconstruct the un-occluded objects regions under sparse view supervision. Our method imposes no constraints on NeRF pre-training, and requires no extra user input or explicit semantic priors. It is an order of magnitude faster than re-training NeRF from scratch while maintaining on-par and even superior performance.

Teaser image showing the results of the NeRF update method.
Figure 1: After training a NeRF for pieces on a Chinese chess board from hundreds of RGB images, a player moves the "Pao" piece, highlighted by the red box, to a new position. To accommodate this physical object reconfiguration, we take 4 additional images for the changed scene, and use our method to estimate scene changes and quickly update the pre-trained NeRF with the guidance of the 4 images.

Results

Pretrained NeRF
Updated NeRF
Retrained NeRF
Pre-trained NeRF
Updated NeRF
Retrained NeRF
Pretrained NeRF
Updated NeRF
Retrained NeRF
Pretrained NeRF
Updated NeRF
Retrained NeRF
Pretrained NeRF
Updated NeRF
Retrained NeRF
Pretrained NeRF
Updated NeRF
Retrained NeRF
Pre-trained NeRF
Updated NeRF
Re-trained NeRF
Pretrained NeRF
Updated NeRF
Retrained NeRF
Pretrained NeRF
Updated NeRF
Retrained NeRF

BibTeX

@article{lu2024fast,
  title = {Fast Sparse View Guided NeRF Update for Object Reconfigurations}, 
  author = {Ziqi Lu and Jianbo Ye and Xiaohan Fei and Xiaolong Li and Jiawei Mo and Ashwin Swaminathan and Stefano Soatto},
  year={2024},
  eprint={2403.11024},
  archivePrefix={arXiv},
  primaryClass={cs.CV}
}