pytorch-enhance: Image Super-Resolution in PyTorch

Library for Minimal Modern Image Super-Resolution in PyTorch

PyTorch Enhance provides a consolidated package of popular Image Super-Resolution models, datasets, and metrics to allow for quick and painless benchmarking or for quickly adding pretrained models to your application.

Documentation

https://pytorch-enhance.readthedocs.io

Installation

pip

pip install torch-enhance

latest

git clone https://github.com/IsaacCorley/pytorch-enhance.git
cd pytorch-enhance
python setup.py install

Models

The following models are currently implemented:

• SRCNN from Dong et. al Image Super-Resolution Using Deep Convolutional Networks
• VDSR from Lee et al. Accurate Image Super-Resolution Using Very Deep Convolutional Networks
• ESPCN from Shi et. al Real-Time Single Image and Video Super-Resolution Using an Efficient Sub-Pixel Convolutional Neural Network
• SRResNet from Ledig et. al Photo-Realistic Single Image Super-Resolution Using a Generative Adversarial Network
• EDSR from Lim et. al Enhanced Deep Residual Networks for Single Image Super-Resolution

import torch
import torch_enhance

# increase resolution by factor of 2 (e.g. 128x128 -> 256x256)
model = torch_enhance.models.SRResNet(scale_factor=2, channels=3)

lr = torch.randn(1, 3, 128, 128)
sr = model(x) # [1, 3, 256, 256]

State-of-the-Art

Not sure which models are currently the best? Check out the PapersWithCode Image Super-Resolution Leaderboards

Datasets

The following benchmark datasets are available for usage:

• BSDS100
• BSDS200
• BSDS300
• BSDS500
• Set5
• Set14
• T91
• Historical
• Urban100
• Manga109
• General100
• DIV2K

Dataset Samples

BSDS300	BSDS500	T91

Set5	Set14	Historical

Losses

• Perceptual Loss (VGG16)

Metrics

• Mean Squared Error (MSE)
• Mean Absolute Error (MAE)
• Peak-Signal-Noise-Ratio (PSNR)

Examples

$ cd examples

• Get up and benchmarking quickly with PyTorch Lightning
• Coming from Keras? Try our example using the Poutyne library

Running Tests

$ pytest -ra

Cite

Please cite this repository if you used this code in your own work:

@software{isaac_corley_2020_3739368,
  author       = {Isaac Corley},
  title        = {PyTorch Enhance},
  month        = apr,
  year         = 2020,
  publisher    = {Zenodo},
  version      = {0.1.2},
  doi          = {10.5281/zenodo.3739368},
  url          = {https://doi.org/10.5281/zenodo.3739368}
}