SR-Caffe

Image Super-Resolution use caffe.

This project implements two articles，"Image Super-Resolution Using Deep Convolutional Networks"(ECCV 2014) and "Accurate Image Super-Resolution Using Very Deep Convolutional Networks" (CVPR 2016 Oral Paper).

工程组织结构

├── dataset       (数据集)
│   ├── hdf5
│   │   ├── SRCNN-5
│   │   ├── SRCNN-Pool
│   │   └── VDSR-20
│   └── images
│       ├── 291
│       ├── 91
│       ├── 91-aug
│       └── test
├── external
│   └── caffe
├── pdf
├── test    （测试评估）
│   ├── images
│   │   ├── Set14
│   │   └── Set5
│   ├── matconvnet
│   ├── results
│   │   ├── SRCNN-5
│   │   ├── SRCNN-Pool
│   │   └── VDSR-20
│   │       ├── official
│   │       └── ours
│   └── utils
└── train   （训练）
    ├── SRCNN-5
    │   ├── caffemodel
    │   ├── convert
    │   ├── demo
    │   │   └── utils
    │   └── logs
    ├── SRCNN-Pool
    │   ├── caffemodel
    │   ├── convert
    │   ├── logs
    │   └── prototxt
    │       └── data_pool
    │           └── logs
    └── VDSR-20
        ├── caffemodel
        ├── convert
        └── logs

运行

1 数据准备

下载数据集 zip(train) zip(test)

git clone https://github.com/mindcont/SR-Caffe.git
cd SR-Caffe && mkdir dataset
wget http://cv.snu.ac.kr/research/VDSR/train_data.zip
wget http://cv.snu.ac.kr/research/VDSR/test_data.zip
unzip train_data.zip images/

# use matlab 
run train/data_aug.m

2 生成HDF5 格式，供caffe 训练

generate_train.m
generate_test.m

3 训练

run train_VDSR-20.sh or train_SRCNN-5.sh

4 测试评估

run demo_SR.m

博客

• 深度学习在图像超分辨率重建中的应用

• 图像超分辨率重构（一）原理及方法总结

类似项目

• "Image super-resolution as sparse representation of raw image patches"(CVPR2008)

• Single Image Super-Resolution from Transformed Self-Exemplars(CVPR2015)

• "Real-Time Single Image and Video Super-Resolution Using an Efficient Sub-Pixel Convolutional Neural Network"(CVPR 2016)

• "Photo-Realistic Single Image Super-Resolution Using a Generative Adversarial Network"(arxiv, 21 Nov, 2016)

• "Deep Laplacian Pyramid Networks for Fast and Accurate Super-Resolution"(CVPR 2017)

• Video-Super-Resolution