Here is the official implementation for CVPR 2022 paper "LAR-SR: A Local Autoregressive Model for Image Super-Resolution".
res_vq.py: textural VQVAE in Stage 1
fold_model.py: LAR-module in Stage 2
fold_attention_model.py: LAR-attn-layer based LAR-module
test.py: test the model with pretrained checkpoint
Due to the size limitation, the checkpoints are not uploaded
If you use this code for your research or project, please cite:
@inproceedings{guo2022lar,
title={Lar-sr: A local autoregressive model for image super-resolution},
author={Guo, Baisong and Zhang, Xiaoyun and Wu, Haoning and Wang, Yu and Zhang, Ya and Wang, Yan-Feng},
booktitle={Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition},
pages={1909--1918},
year={2022}
}