PyTorch ByNet

Implementation of ICIP2017 Paper: "ByNet-SR: Image Super Resolution with a Bypass Connection Network " in PyTorch

Citation

If you find the code and datasets useful in your research, please cite:

@InProceedings{ByNet,
    author    = {Xu, J. and Chae, Y. and Stenger, B.}, 
    title     = {{ByNet-SR}: Image Super Resolution with a Bypass Connection Network}, 
    booktitle = {IEEE International Conference on Image Processing},
    year      = {2017}
}

Usage

Training

usage: main_bynet.py [-h] [--batchSize BATCHSIZE] [--nEpochs NEPOCHS] [--lr LR]
               [--step STEP] [--cuda] [--resume RESUME]
               [--start-epoch START_EPOCH] [--clip CLIP] [--threads THREADS]
               [--momentum MOMENTUM] [--weight-decay WEIGHT_DECAY]
               [--pretrained PRETRAINED]

optional arguments:
  -h, --help            show this help message and exit
  --batchSize BATCHSIZE
                        Training batch size
  --nEpochs NEPOCHS     Number of epochs to train
  --lr LR               Init learning Rate. Default=0.1
  --step STEP           Sets the learning rate to the initial LR decayed by
                        momentum every n epochs, Default: n=10
  --cuda                use cuda?
  --resume RESUME       Path to latest checkpoint (default: none)
  --start-epoch START_EPOCH
                        Manual epoch number (useful on restarts)
  --clip CLIP           Clipping Gradients. Default=0.5
  --threads THREADS     Number of threads for data loader to use
  --momentum MOMENTUM   Momentum
  --weight-decay WEIGHT_DECAY, --wd WEIGHT_DECAY
                        weight decay, Default: 1e-4

An example of training usage is shown as follows:

python main_bynet.py --cuda

Evaluation

usage: eval.py [-h] [--cuda] [--model MODEL] [--dataset DATASET]
               [--scale SCALE]

PyTorch ByNet Eval

optional arguments:
  -h, --help         Show this help message and exit
  --cuda             use cuda?
  --model MODEL      Model path. Default=model/model_epoch_40.pth
  --dataset DATASET  Dataset name, Default: Set5

An example of training usage is shown as follows:

python eval.py --cuda

Demo

usage: demo.py [-h] [--cuda] [--model MODEL] [--image IMAGE] [--scale SCALE]
               
optional arguments:
  -h, --help            Show this help message and exit
  --cuda                Use cuda
  --model               Model path. Default=model/model_epoch_40.pth
  --image               Image name. Default=butterfly_GT
  --scale               Scale factor, Default: 4

An example of demo usage is shown as follows:

python demo.py --cuda

Image Process without Matlab

usage: run_without_matlab.py [-h] [--cuda] [--model MODEL] [--folder FOLDER]

PyTorch ByNet Enhance (without matlab)

optional arguments:
  -h, --help       show this help message and exit
  --cuda           use cuda?
  --model MODEL    Model path, Default=model/model_epoch_40.pth
  --folder FOLDER  Folder name

An example of demo usage is shown as follows:

python run_without_matlab.py --cuda --folder Set5

Put all images in the folder you defined and the results will be appeared in result folder

Prepare Training dataset

• We provide a simple hdf5 format training sample in data folder with 'data' and 'label' keys, the training data is generated with Matlab Bicubic Interplotation, please refer Code for Data Generation for creating training files.

Performance

• We provide a pretrained ByNet9 model trained on 291 images with data augmentation
• No bias is used in this implementation
• Performance in PSNR on Set5

Scale	PNSR
2x	37.75
3x	33.96
4x	31.60

Result

From left to right are ground-truth image, bicubic and ByNet