Enhanced Adaptive Dense Connection Single Image Super-Resolution
Abstract
Increasing model size often results in improved performance on super-resolution reconstruction. However, at some point large model cannot SR huge images due to GPU/TPU memory limitations. In this paper, to address this problem, we present Block-Reconstruction(BR) strategy to improve the reconstruction quality of large images, which lower memory consumption. Meanwhile, we propose an enhanced adaptive dense connection super resolution reconstruction network(EDCSR) that has 89M parameters. In AIM2020 Real Image Super-Resolution Challenge, we won the second place in Track 1 and Track 2, and the third place in Track 3.
Cite
Text
Xie et al. "Enhanced Adaptive Dense Connection Single Image Super-Resolution." European Conference on Computer Vision Workshops, 2020. doi:10.1007/978-3-030-67070-2_26Markdown
[Xie et al. "Enhanced Adaptive Dense Connection Single Image Super-Resolution." European Conference on Computer Vision Workshops, 2020.](https://mlanthology.org/eccvw/2020/xie2020eccvw-enhanced/) doi:10.1007/978-3-030-67070-2_26BibTeX
@inproceedings{xie2020eccvw-enhanced,
title = {{Enhanced Adaptive Dense Connection Single Image Super-Resolution}},
author = {Xie, Tangxin and Li, Jing and Shen, Yi and Jia, Yu and Zhang, Jialiang and Zeng, Bing},
booktitle = {European Conference on Computer Vision Workshops},
year = {2020},
pages = {437-452},
doi = {10.1007/978-3-030-67070-2_26},
url = {https://mlanthology.org/eccvw/2020/xie2020eccvw-enhanced/}
}