egonerf's Introduction

Balanced Spherical Grid for Egocentric View Synthesis

Official implementation of Balanced Spherical Grid for Egocentric View Synthesis (CVPR 2023)

TL;DR - One Sentence Summary

EgoNeRF is a grid-based NeRF model that utilize a balanced spherical grid to reconstruct large-scale egocentric captured scenes.

Installation

Environment Setup

We tested our code on Ubuntu 20.04 with RTX 3090 GPU. With proper version of CUDA toolkit, it would work on other environments.

conda create -n EgoNeRF python=3.8
conda activate EgoNeRF
pip install torch torchvision --extra-index-url https://download.pytorch.org/whl/cu113
pip install -r requirements.txt

Dataset

We provide OmniBlender and Ricoh360 datset. You can download the dataset from the google drive link above. Put the data in the directory data/

Training EgoNeRF

To train EgoNeRF, run the scripts below.

python train.py --config configs/omniblender/barbershop/default.txt

Citation

Cite as below if you find this paper and repository are helpful to you:

@InProceedings{Choi_2023_CVPR,
    author    = {Choi, Changwoon and Kim, Sang Min and Kim, Young Min},
    title     = {Balanced Spherical Grid for Egocentric View Synthesis},
    booktitle = {Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR)},
    month     = {June},
    year      = {2023},
    pages     = {16590-16599}
}

Acknowledgement

This repository borrows codes from the amazing work TensoRF.

egonerf's People

Contributors

Stargazers

Watchers

egonerf's Issues

Hi! I have read the paper and consider EgoNeRF an incredible equirectangular image reconstruction method in terms of training speed and novel-view synthesis quality. I would like further seeing the detailed rendering FPS data to consider real-time applications.

Thank you!

changwoonchoi / egonerf Goto Github PK