DeblurGAN-v2
We present a new end-to-end generative adversarial network (GAN) for single image motion deblurring, named DeblurGAN-v2, which considerably boosts state-of-the-art deblurring efficiency, quality, and flexibility. DeblurGAN-v2 is based on a relativistic conditional GAN with a double-scale discriminator. For the first time, we introduce the Feature Pyramid Network into deblurring, as a core building block in the generator of DeblurGAN-v2. It can flexibly work with a wide range of backbones, to navigate the balance between performance and efficiency. The plug-in of sophisticated backbones (e.g., Inception-ResNet-v2) can lead to solid state-of-the-art deblurring. Meanwhile, with light-weight backbones (e.g., MobileNet and its variants), DeblurGAN-v2 reaches 10-100 times faster than the nearest competitors, while maintaining close to state-of-the-art results, implying the option of real-time video deblurring. We demonstrate that DeblurGAN-v2 obtains very competitive performance on several popular benchmarks, in terms of deblurring quality (both objective and subjective), as well as efficiency. Besides, we show the architecture to be effective for general image restoration tasks too.
DeblurGAN-v2 Architecture
Datasets
The datasets for training can be downloaded via the links below:
Training
Command
python train.py
training script will load config under config/config.yaml
Tensorboard visualization
Pre-trained models
Dataset | G Model | D Model | Loss Type | PSNR/ SSIM | Link |
---|---|---|---|---|---|
GoPro Test Dataset | InceptionResNet-v2 | double_gan | ragan-ls | 29.55/ 0.934 | https://drive.google.com/open?id=1hv238vS9VeK8C8QwrQ2MnJc3ChgHyaGy |
MobileNet | double_gan | ragan-ls | 28.17/ 0.925 | https://drive.google.com/open?id=1JhnT4BBeKBBSLqTo6UsJ13HeBXevarrU | |
MobileNet-DSC | double_gan | ragan-ls | 28.03/ 0.922 |
Citation
If you use this code for your research, please cite our paper.
```
@InProceedings{Kupyn_2019_ICCV,
author = {Orest Kupyn and Tetiana Martyniuk and Junru Wu and Zhangyang Wang},
title = {DeblurGAN-v2: Deblurring (Orders-of-Magnitude) Faster and Better},
booktitle = {The IEEE International Conference on Computer Vision (ICCV)},
month = {Oct},
year = {2019}
}
```