This is the code repository for the ECCV 2022 submission "Neural Scene Decoration from a Single Photograph".
- Download the Structured3D dataset and organize it in the following format:
Structured3D
- data
- scene_00000
- scene_00001
- ...
- scene_03499
- splits
- bedroom_obj
- living_room_obj
where splits
are provided in this repository, and the scenes in data
are downloaded from Structured3D.
- Create virtual environment and install the dependencies. By default, the CUDA version used is 11.3.
conda create --name nsd python=3.8
pip install -r requirements.txt
-
Change dataset path (
--data
) intrain_struct3d.sh
to -
Run
train_struct3d.sh
to start training.
This respository is modified from https://github.com/lucidrains/lightweight-gan.
Please cite our paper if you find it useful for your research:
@inproceedings{pang2022nsd,
title={Neural Scene Decoration from a Single Photograph},
author={Hong-Wing Pang and Yingshu Chen and Phuoc-Hieu Le and Binh-Son Hua and Duc Thanh Nguyen and Sai-Kit Yeung},
year={2022},
booktitle={ECCV},
}