Super-resolution

some cross-project repo: Martin Krasser/Keras-super-resolution: SRGAN, EDSR, WDSR
open-mmlab/MMSR: SRResNet, SRGAN, ESRGAN, EDVR, etc.

SRCNN (TPAMI 2014)

Image Super-Resolution Using Deep Convolutional Networks by CUHK
Project - include Matlab and Caffe code
first upscale it to the desired size using bicubic interpolation

FSRCNN (ECCV 2016)

Accelerating the Super-Resolution Convolutional Neural Network by CUHK
use deconvolution to upscale instead of pre-processing

VDSR (CVPR 2016)

Accurate Image Super-Resolution Using Very Deep Convolutional Networks
start to use residual, one long skip-connection
../../_images/VDSR.png
Same as SRCNN, it take upscaled LR image as imput

SRGAN (CVPR 2017)

Photo-Realistic Single Image Super-Resolution Using a Generative Adversarial Network by Twitter
Tensorflow 2.0
SRResNet + long skip-connection, diverge from MSE novel perceptual loss using high-level feature maps of VGG network

EDSR (CVPR 2017)

Enhanced Deep Residual Networks for Single Image Super-Resolution remove batch normalization of SRResNet→ save resource spend resource on convolution instead

WDSR (CVPR 2018)

Wide Activation for Efficient and Accurate Image Super-Resolution
pyTorch based on EDSR | tensorflow ../../_images/WDSR_block.png
increase channel before ReLU linear low-rank convolution stack and large channel Official implemenetation (PyTorch based on EDSR) TensorFlow-WDSR (official) Keras-super-resolution (Single Image Super-Resolution with WDSR, EDSR and SRGAN) By Martin Krasser

SFTGAN (CVPR 2018)

Recovering Realistic Texture in Image Super-resolution by Deep Spatial Feature Transform by senseTime
Spatial Feature Transform: learns a mapping function \(M\) that outputs a modulation parameter pair (\( \) generate affine transformation parameters for spatial-wise feature modulation
../../_images/SFTGAN.png ../../_images/SFTGAN_architecture.png

ESRGAN (ECCV 2018)

ESRGAN: Enhanced Super-Resolution Generative Adversarial Networks from CUHK
RRDB (Residual in Residual Dense Blcok)