ofa-for-super-resolution
所属分类:论文
开发工具:Python
文件大小:0KB
下载次数:0
上传日期:2021-07-01 07:30:28
上 传 者:
sh-1993
说明: 基于“一劳永逸:训练一个网络并将其专门化以实现高效部署”(ICLR...,
(Image downscaling & super-resolution project based on "Once for All: Train One Network and Specialize it for Efficient Deployment" (ICLR 2020))
文件列表:
LICENSE (11346, 2021-07-01)
build.sh (106, 2021-07-01)
eval_ofa_net.py (2427, 2021-07-01)
eval_ofa_net_sr.py (9932, 2021-07-01)
eval_specialized_net.py (6519, 2021-07-01)
figures/ (0, 2021-07-01)
figures/cnn_imagenet_new.png (807351, 2021-07-01)
figures/diverse_hardware.png (805688, 2021-07-01)
figures/imagenet_80_acc.png (550009, 2021-07-01)
figures/ofa-tutorial.jpg (131312, 2021-07-01)
figures/overview.png (594591, 2021-07-01)
figures/video_figure.png (294316, 2021-07-01)
img_car2.png (166254, 2021-07-01)
img_car4.png (177864, 2021-07-01)
img_supernet.png (105768, 2021-07-01)
independent/ (0, 2021-07-01)
independent/color_histogram_difference.py (1001, 2021-07-01)
independent/crop_and_save.py (572, 2021-07-01)
independent/mp4_to_png.py (1607, 2021-07-01)
independent/resize_and_save.py (481, 2021-07-01)
independent/uvg_to_png.py (5751, 2021-07-01)
ofa/ (0, 2021-07-01)
ofa/__init__.py (0, 2021-07-01)
ofa/elastic_nn/ (0, 2021-07-01)
ofa/elastic_nn/__init__.py (0, 2021-07-01)
ofa/elastic_nn/modules/ (0, 2021-07-01)
ofa/elastic_nn/modules/__init__.py (0, 2021-07-01)
ofa/elastic_nn/modules/dynamic_layers.py (13601, 2021-07-01)
ofa/elastic_nn/modules/dynamic_op.py (8379, 2021-07-01)
ofa/elastic_nn/networks/ (0, 2021-07-01)
ofa/elastic_nn/networks/__init__.py (364, 2021-07-01)
ofa/elastic_nn/networks/ofa_mbs4.py (24499, 2021-07-01)
ofa/elastic_nn/networks/ofa_mbv3.py (17318, 2021-07-01)
ofa/elastic_nn/networks/ofa_mbx4.py (29474, 2021-07-01)
ofa/elastic_nn/networks/ofa_proxyless.py (16134, 2021-07-01)
ofa/elastic_nn/training/ (0, 2021-07-01)
ofa/elastic_nn/training/__init__.py (0, 2021-07-01)
ofa/elastic_nn/training/progressive_shrinking.py (23405, 2021-07-01)
... ...
# Image Downscaling & Super-Resolution based on Once-for-All
This repository contains image downscaling & super-resolution project code based on the paper ["Once-for-All: Train One Network and Specialize it for Efficient Deployment"](https://arxiv.org/abs/1908.09791) (ICLR 2020).
The objectives of this proejct are
* Find the best image downscaling & super-resolution neural network architecture on mobile devices
* Support both 2x, 4x super-resolution in a single architecture.
## License and Citation
```BibTex
@inproceedings{
cai2020once,
title={Once for All: Train One Network and Specialize it for Efficient Deployment},
author={Han Cai and Chuang Gan and Tianzhe Wang and Zhekai Zhang and Song Han},
booktitle={International Conference on Learning Representations},
year={2020},
url={https://arxiv.org/pdf/1908.09791.pdf}
}
```
```BibTex
@inproceedings{
kim2018tar,
title={Task-Aware Image Downscaling},
author={Heewon Kim and Myungsub Choi and Bee Lim and Kyoung Mu Lee},
booktitle={European Conference on Computer Vision},
year={2018},
url={https://openaccess.thecvf.com/content_ECCV_2018/papers/Heewon_Kim_Task-Aware_Image_Downscaling_ECCV_2018_paper.pdf}
}
```
# 2x/4x Image Downscaling & Super-Resolution in a Single Mobile Architecture
#### Overview of Supernet Architecture
![img](img_supernet.png)
#### Progressive Shrinking of
* Kernel Size
* Network Depth
* Expand Ratio
* Number of Pixelshuffle
### Comparison to CAR in terms of PSNR
["CAR: Learned Image Downscaling for Upscaling using Content Adaptive Resampler"](https://arxiv.org/abs/1907.12904)
Dataset |
Ours |
CAR |
Set14-2xUP |
39.15 |
35.61 |
Set14-4xUP |
31.01 |
30.30 |
![img](img_car2.png)
![img](img_car4.png)
近期下载者:
相关文件:
收藏者: