DFDNet: A Python repository from Jahidur1414

Blind Face Restoration via Deep Multi-scale Component Dictionaries

Note: This 'master' branch only contains the face region without putting them into the origial input. You can refer to 'whole' branch for flexible restoration.

Overview of our proposed method. It mainly contains two parts: (a) the off-line generation of multi-scale component dictionaries from large amounts of high-quality images, which have diverse poses and expressions. K-means is adopted to generate K clusters for each component (i.e., left/right eyes, nose and mouth) on different feature scales. (b) The restoration process and dictionary feature transfer (DFT) block that are utilized to provide the reference details in a progressive manner. Here, DFT-i block takes the Scale-i component dictionaries for reference in the same feature level.

(a) Offline generation of multi-scale component dictionaries.

(b) Architecture of our DFDNet for dictionary feature transfer.

Pre-train Models and dictionaries

Downloading from the following url and put them into ./.

BaiduNetDisk (s9ht)
GoogleDrive

These folder structure should be:

.
├── checkpoints                    
│   ├── facefh_dictionary                  
│   │   └── latest_net_G.pth   
├── weights
│   └── vgg19.pth
├── DictionaryCenter512
│   ├── right_eye_256_center.npy
│   ├── right_eye_128_center.npy
│   ├── right_eye_64_center.npy
│   ├── right_eye_32_center.npy
│   └── ...
└── ...

Prerequisites

Pytorch (≥1.1 is recommended)
dlib

face-alignment

cd ./FaceLandmarkDetection
python setup.py install
cd ..

Testing

Crop face from the whole image.

cd ./CropFace
python crop_face_dlib.py

(You can change the image path and save path in line 61~62)

Compute the facial landmarks.

cd ./FaceLandmarkDetection
python get_face_landmark.py

This code is mainly borrowed from this work. You can change the image path and save path in line 17~18. )_

Run the face restoration.

python test_FaceDict.py

（You can directly run this code for the given test images and landmarks without step 1 and 2. The image path can be revised in line 100~103.）

Some plausible restoration results on real low-quality images

Input	Results

Citation

@InProceedings{Li_2020_ECCV,
author = {Li, Xiaoming and Chen, Chaofeng and Zhou, Shangchen and Lin, Xianhui, Zuo, Wangmeng and Zhang, Lei},
title = {Blind Face Restoration via Deep Multi-scale Component Dictionaries},
booktitle = {ECCV},
year = {2020}
}

This work is licensed under a Creative Commons Attribution-NonCommercial-ShareAlike 4.0 International License.

Jahidur1414/DFDNet