MRDNet

Image deblurring method based on self-attention and residual wavelet transform

Contents

The directory for this readme document is as follows：
1.Dependencies and Installation
2.Datasets Preparation
3.Model
4.Train Ways
5.Test Ways

---

1.Dependencies and Installation

• Ubuntu 16.04
• Python 3.8
• pytorch1.7
• cuda11.0
• Scikit-image
• opencv-python
• Tensorboard

2.Datasets Preparation

• Download GoPro datasets from Baidu Netdisk https://pan.baidu.com/s/1wtR5gIZBoW-aYSx447ZPRQ (the password is 8k2q)
• Copy the downloaded dataset to the dataset folder

3.Model

• Download pretrained model from Baidu Netdisk https://pan.baidu.com/s/1sueSRt1MjSpImP-zVBX4gA (the password is whui)
• Load the pretrained model into the project folder and place the downloaded model in the weights folder

4.Train Ways

python main.py --model_name MRDNet --mode train --data_dir dataset/GOPRO

5.Test Ways

python main.py --model_name MRDNet --mode test --data_dir dataset/GOPRO --test_model weights/MRDNet.pkl