本仓库将对一些论文中提出的SR算法进行总结和复现，使用的推理框架是TensorFlow。
Github文档地址：https://lankning.github.io/Super-Resolution/
- SRCNN 2014
- FSRCNN 2016
- ESPCN 2016
- VESPCN 2017 （Only Notes）
- DUF 2018
- FALSR 2019（Only Notes）
- TGA 2020（Only Notes）
- One-Stage STVSR（Partly codes）
- FSR 2021
tensorFlow-gpu>=2.2, <=2.5
jupyter notebook
opencv-python
pillow
matplotlib
- 如果你有conda，建议使用conda创建一个新的环境，并安装
TensorFlow等必要的库
conda create -n tf22 python=3.7
conda activate tf22
conda install tensorflow-gpu=2.2
pip install -r requirements.txt -i https://pypi.mirrors.ustc.edu.cn/simple/- 使用
jupyter运行data_prepare.ipynb文件，在本文件夹打开终端，输入
jupyter notebook- 以
SRCNN为例，运行SRCNN.ipynb文件是重新训练模型，运行LR2HR.ipynb文件是将低清视频通过模型转为高清。仍然在jupyter中打开，运行。
- data
- train_data
- train_data?x
- demo.mp4 (you can require this demo video by https://wwe.lanzous.com/iXi1dow5r3g)
- Re-Implementated Model
- model
- model.h5
- static
- paper.pdf
- images.png
- SRCNN.ipynb: train SRCNN model
- LR2HR.ipynb: transfer LR frames to HR video using trained model
- utils.ipynb: ipynb version of utils
- utils.py: python module for use
- README.md
- model
- data_prepare.ipynb: prepare data from video
- utils.py: python module for use
- README.md
- requirements.txt
- get frame from video
- transfer HR images to LR images
- Low-resolution: eg. ./data/train_data4x/: The scaled frames by 4x which will be used in train stage eg. ./data/train_data2x/: The scaled frames by 2x which will be used in train stage
- Original-resolution: ./data/train_data/: The HR frames which will be used in train stage coresponde to ./data/x_train_data?x/
朱文康
如果你有问题，欢迎联系我。我的邮箱是：wenkang_zhu@whu.edu.cn，乐意回复。
谢谢。