A Simple Plugin for Transforming Images to Arbitrary Scales

This repository contains the official PyTorch implementation of ARIS: https://arxiv.org/abs/2210.03417.

Requirements

A suitable conda environment named aris can be created and activated with:

conda env create -f environment.yaml
conda activate aris

Model Zoo

https://drive.google.com/drive/folders/1GmZiTC5NSx-UnT0SwPeJQqjWtThc6yh-?usp=sharing

Train

Comming soon....

Test

Test set: https://drive.google.com/drive/folders/1GmZiTC5NSx-UnT0SwPeJQqjWtThc6yh-?usp=sharing

python test_final.py --model hat+aris \
--dir_data 'xxx/dataset' \
--pretrain 'xxx/HAT_ARIS.pt' \
--data_test Set5+Set14+B100+Urban100(or DIV2K) \
--scale 2+3+4+6+8 \
--n_GPUs 1 \
--save 'xxx/xxx' \
--crop_batch_size 1 \
--num_heads 6 \
--num_layers 6 

Here we release the checkpoint of 'HAT+ARIS' with 6 heads and 6 layers, more checkpoints will comming soon.

Citation

If you use this code for your research or project, please cite:

@article{zhou2022aris,
author    = {Zhou, Qinye and Li, Ziyi and Xie, Weidi and Zhang, Xiaoyun and Zhang, Ya and Wang, Yanfeng},
title     = {A Simple Plugin for Transforming Images to Arbitrary Scales},
booktitle = {British Machine Vision Conference (BMVC)}，
year      = {2022}
}

Acknowledgements

Many thanks to the code bases from IPT, HAT, SwinIR.