SimDeblur

SimDeblur (Simple Deblurring) is an open source framework for image and video deblurring toolbox based on PyTorch, which contains most deep-learning based state-of-the-art deblurring algorithms. It is easy to implement your own image or video deblurring or other restoration algorithms.

Major features

• Modular Design

The toolbox decomposes the deblurring framework into different components and one can easily construct a customized restoration framework by combining different modules.

• State of the art

The toolbox contains most deep-learning based state-of-the-art deblurring algorithms, including MSCNN, SRN, DeblurGAN, EDVR, etc.

• Distributed Training

New Features

[2021/3/31] support DVD, GoPro and REDS video deblurring datasets. [2021/3/21] first release.

Surpported Methods and Benchmarks

• Single Image Deblurring

  • MSCNN [Paper, Project]
  • SRN [Paper, Project]
  • DeblurGAN [Paper, Project]
  • DMPHN [Paper, Project]
  • DeblurGAN_V2 [Paper, Project]
  • SAPHN [Paper]

• Video Deblurring

  • DBN [Paper, Project]
  • STRCNN [paper]
  • DBLRNet [Paper]
  • EDVR [Paper, Project]
  • STFAN [Paper, Project]
  • IFIRNN [Paper]
  • CDVD-TSP [Paper, Project]
  • ESTRNN [Paper, Project]

• Benchmarks

  • GoPro [Paper, Data]
  • DVD [Paper, Data]
  • REDS [Paper, Data]

Dependencies and Installation

• Python 3 (Conda is recommended)
• Pytorch 1.5.1 (with GPU)
• CUDA 10.2+

1. Clone the repositry or download the zip file

    git clone https://github.com/ljzycmd/SimDeblur.git
   

2. Install SimDeblur

   # create a pytorch env
   conda create -n simdeblur python=3.7
   conda activate simdeblur   
   # install the packages
   cd SimDeblur
   bash Install.sh

Usage

1 Start with trainer

You can construct a simple training process use the default trainer like following:

from simdeblur.config import build_config, merge_args
from simdeblur.engine.parse_arguments import parse_arguments
from simdeblur.engine.trainer import Trainer


args = parse_arguments()

cfg = build_config(args.config_file)
cfg = merge_args(cfg, args)
cfg.args = args

trainer = Trainer(cfg)
trainer.train()

Then start training with single GPU:

CUDA_VISIBLE_DEVICES=0 bash ./tools/train.sh ./config/dbn/dbn_dvd.yaml 1

multi GPU training:

CUDA_VISIBLE_DEVICES=0,1,2,3 bash ./tools/train.sh ./config/dbn/dbn_dvd.yaml 4

2 Build each module

The SimDeblur also provides you to build each module. build the a dataset:

from easydict import EasyDict as edict
from simdeblur.dataset import build_dataset

dataset = build_dataset(edict({
    "name": "DVD",
    "mode": "train",
    "sampling": "n_c",
    "overlapping": True,
    "interval": 1,
    "root_gt": "./dataset/DVD/quantitative_datasets",
    "num_frames": 5,
    "augmentation": {
        "RandomCrop": {
            "size": [256, 256] },
        "RandomHorizontalFlip": {
            "p": 0.5 },
        "RandomVerticalFlip": {
            "p": 0.5 },
        "RandomRotation90": {
            "p": 0.5 },
    }
}))

print(dataset[0])

build the model:

from simdeblur.model import build_backbone

model = build_backbone({
    "name": "DBN",
    "num_frames": 5,
    "in_channels": 3,
    "inner_channels": 64
})

x = torch.randn(1, 5, 3, 256, 256)
out = model(x)

build the loss:

from simdeblur.model import build_loss

criterion = build_loss({
    "name": "MSELoss",
})
x = torch.randn(2, 3, 256, 256)
y = torch.randn(2, 3, 256, 256)
print(criterion(x, y))

And the optimizer and lr_scheduler also can be created by "build_optimizer" and "build_lr_scheduler" etc.

Dataset Description

Click here for more information.

Acknowledgment

[1] facebookresearch. detectron2. https://github.com/facebookresearch/detectron2

[2] subeeshvasu. Awesome-Deblurring. https://github.com/subeeshvasu/Awesome-Deblurring

Citations

If SimDeblur helps your research or work, please consider citing SimDeblur.

@misc{cao2021simdeblur,
  author =       {Mingdeng Cao},
  title =        {SimDeblur},
  howpublished = {\url{https://github.com/ljzycmd/SimDeblur}},
  year =         {2021}
}

If you have any question, please contact me at mingdengcao AT gmail.com.