Skip to content

Official repository of ActiveNeRF (ECCV2022)

Notifications You must be signed in to change notification settings

abubake/ActiveNeRF

 
 

Repository files navigation

ActiveNeRF: Learning where to See with Uncertainty Estimation (ActiveNeRF)

This repo contains the official PyTorch code for ActiveNeRF [paper].

Introduction

main

We present a novel learning framework, ActiveNeRF, aiming to model a 3D scene with a constrained input budget. We first incorporate uncertainty estimation into a NeRF model, which ensures robustness under few observations and provides an interpretation of how NeRF understands the scene. On this basis, we propose to supplement the existing training set with newly captured samples based on an active learning scheme. By evaluating the reduction of uncertainty given new inputs, we select the samples that bring the most information gain. In this way, the quality of novel view synthesis can be improved with minimal additional resources.

Visualizations

Installation

git clone https://github.com/abubake/ActiveNeRF.git
cd ActiveNeRF

We provide a conda environment setup file as analterantive to installing dependencies via requirements.txt. Create the conda environment activenerf by running:

conda env create -f environment.yml
conda activate activenerf

OR

pip install -r requirements.txt

Quick Start

Download data for example dataset: lego

bash download_example_data.sh

Then move "transforms_holdout.json" from "./data/lego" into "./data/nerf_synthetic/lego" which is created when data is downloaded.

Train ActiveNeRF:

python run_nerf.py --config configs/lego_active.txt --expname active_lego --datadir ./data/nerf_synthetic/lego

Contact

If you have any question, please feel free to contact the authors. Xuran Pan: [email protected].

Acknowledgement

Our code is based on NeRF-Pytorch, and NeRF-Tensorflow.

Citation

If you find our work is useful in your research, please consider citing:

@inproceedings{pan2022activenerf,
  title={ActiveNeRF: Learning Where to See with Uncertainty Estimation},
  author={Pan, Xuran and Lai, Zihang and Song, Shiji and Huang, Gao},
  booktitle={Computer Vision--ECCV 2022: 17th European Conference, Tel Aviv, Israel, October 23--27, 2022, Proceedings, Part XXXIII},
  pages={230--246},
  year={2022},
  organization={Springer}
}

About

Official repository of ActiveNeRF (ECCV2022)

Resources

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published

Languages

  • Python 99.0%
  • Shell 1.0%