raoumer / srrescycgan

Deep Cyclic Generative Adversarial Residual Convolutional Networks for Real Image Super-Resolution

  • Public
  • 131 runs
  • GitHub
  • Paper
  • License
Run with an API
Playground API Examples README Versions

😵 Uh oh! This model can't be run on Replicate because it was built with a version of Cog or Python that is no longer supported. Consider opening an issue on the model's GitHub repository to see if it can be updated to use a recent version of Cog. If you need any help, please Contact us about it.

Run time and cost

This model runs on CPU hardware. We don't yet have enough runs of this model to provide performance information.

Readme

Intelligent image scaling to 4x resolution.

Examples

Input Output

Usage

Given that you have a folder of low-resolution images in the folder ./input, the following command saves high-resolution results to the folder ./output.

Run on GPU

This model requires an NVIDIA GPU, compatible with CUDA 11.0.

docker run --gpus=all -it \
    -v $PWD/srrescycgan_code_demo/samples:/input \
    -v $PWD/output:/output \
    us-docker.pkg.dev/replicate/raoumer/srrescycgan:gpu \
    --input-folder=/input \
    --output-folder=/output

Run on CPU

docker run -it \
    -v $PWD/srrescycgan_code_demo/samples:/input \
    -v $PWD/output:/output \
    us-docker.pkg.dev/replicate/raoumer/srrescycgan:cpu \
    --input-folder=/input \
    --output-folder=/output

Arguments

  • --model - Model variant to use. Options:
    • jpeg-compression
    • real-image-corruptions
    • sensor-noise
    • unknown-compressions (default)
  • --no-chop - Don’t chop the image (uses more memory)

Abstract

Recent deep learning based single image super-resolution (SISR) methods mostly train their models in a clean data domain where the low-resolution (LR) and the high-resolution (HR) images come from noise-free settings (same domain) due to the bicubic down-sampling assumption. However, such degradation process is not available in real-world settings. We consider a deep cyclic network structure to maintain the domain consistency between the LR and HR data distributions, which is inspired by the recent success of CycleGAN in the image-to-image translation applications. We propose the Super-Resolution Residual Cyclic Generative Adversarial Network (SRResCycGAN) by training with a generative adversarial network (GAN) framework for the LR to HR domain translation in an end-to-end manner. We demonstrate our proposed approach in the quantitative and qualitative experiments that generalize well to the real image super-resolution and it is easy to deploy for the mobile/embedded devices. In addition, our SR results on the AIM 2020 Real Image SR Challenge datasets demonstrate that the proposed SR approach achieves comparable results as the other state-of-art methods.

Video demo

SRResCycGAN Architecture

Overall Representative diagram

Quantitative Results

The x4 SR quantitative results comparison of our method with others over the DIV2K validation-set (100 images). The best performance is shown in red and the second best performance is shown in blue.

Visual Results

DIV2K Validation-set (100 images)

Here are the SR results comparison of our method on the DIV2K validation-set images.

Real-Image SR Challenge dataset images (Track-3)

Validation-set

You can download all the SR results of our method on the AIM 2020 Real-Image SR validation-set from the Google Drive: SRResCycGAN.

Test-set

You can download all the SR results of our method on the AIM 2020 Real-Image SR test-set from the Google Drive: SRResCycGAN.