We propose a novel multi-scale residual network (MSRN) to fully exploit the image features, which performance exceeds most of the state-of-the-art SR methods. Based on the residual block, we introduce convolution kernels of different sizes to adaptively detect the image features at different scales. Meanwhile, we let these features interact with each other to get the most effective image information. This structure is called Multi-scale Residual Block (MSRB), which effectively extracts the features of the image. Furthermore, the outputs of each MSRB are used as the hierarchical features for global feature fusion. And then, all these features are sent to the reconstruction module for recovery of the high-quality image.

Paper can be found at http://openaccess.thecvf.com/content_ECCV_2018/papers/Juncheng_Li_Multi-scale_Residual_Network_ECCV_2018_paper.pdf

More SR images reconstructed by our model can be found at https://goo.gl/bGnZ8D.

1. Linux
2. Python 3.5
3. PyTorch 0.4.0
4. CPU or NVIDIA GPU + CUDA CuDNN (CUDA 8.0)

python main.py --cuda

python test.py --cuda   

We use matlab code for training data augment

  cd Data_process
  run generate_train.m

  cd Data_process
  run generate_test.m

 cd example
 python test.py --cuda