Deblurring Face Images using Uncertainty Guided Multi-Stream Semantic Networks
Rajeev Yasarla, Federico Perazzi, Vishal M. Patel
Paper Link(IEEE TIP'20) arxiv version
@ARTICLE{ryasarla_UMSN,
author={R. {Yasarla} and F. {Perazzi} and V. M. {Patel}},
journal={IEEE Transactions on Image Processing},
title={Deblurring Face Images Using Uncertainty Guided Multi-Stream Semantic Networks},
year={2020},
volume={29},
number={},
pages={6251-6263},}
We propose a novel multi-stream architecture and training methodology that exploits semantic labels for facial image deblurring. The proposed Uncertainty Guided MultiStream Semantic Network (UMSN) processes regions belonging to each semantic class independently and learns to combine their outputs into the final deblurred result. Pixel-wise semantic labels are obtained using a segmentation network. A predicted confidence measure is used during training to guide the network towards challenging regions of the human face such as the eyes and nose. The entire network is trained in an endto-end fashion.
pip install -r requirements.txt
https://sites.google.com/site/ziyishenmi/cvpr18_face_deblur
python test_face_deblur.py --dataroot ./facades/github/ --valDataroot <path_to_test_data> --netG ./pretrained_models/Deblur_epoch_Best.pth
python train_face_deblur.py --dataroot <path_to_train_data> --valDataroot ./facades/github/ --exp ./face_deblur --batchSize 10
Train Segmentation Netweork using the following command
python seg_train.py --modeclean 1 --dataroot ./facades/github/ --valDataroot ./facades/github/
python seg_train.py --modeclean 0 --dataroot ./facades/github/ --valDataroot ./facades/github/