Efficient Face Super-Resolution via Wavelet-based Feature Enhancement Network (ACMMM 2024)

Paper (Arxiv) | Supplementary Material | [Project Page]()

Installation and Requirements

I have trained and tested the codes on

• Ubuntu 20.04
• CUDA 11.1
• Python 3.8, install required packages by pip install -r requirements.txt

Getting Started

Download Our Pretrain Models and Test Dataset. Additionally, we offer our FSR results in orginal paper.

Note：Test results are slightly different from the original paper because the model weights were obtained by re-training after organizing our codes.

• Pretrain_Models
• Test_Datasets
• FSR_Results_in_Orginal_Paper

Test with Pretrained Models

# On CelebA Test set
python test.py --gpus 1 --model wfen --name wfen \
    --load_size 128 --dataset_name single --dataroot /path/to/datasets/test_datasets/CelebA1000/LR_x8_up/ \
    --pretrain_model_path ./pretrain_models/wfen/wfen_best.pth \
    --save_as_dir results_celeba/wfen

# On Helen Test set
python test.py --gpus 1 --model wfen --name wfen \
    --load_size 128 --dataset_name single --dataroot /path/to/datasets/test_datasets/Helen50/LR_x8_up/ \
    --pretrain_model_path ./pretrain_models/wfen/wfen_best.pth \
    --save_as_dir results_helen/wfen

Evaluation

We provide evaluation codes in script test.sh for calculate PSNR/SSIM/LPIPS/VIF/Parmas/FLOPs scores.

Train the Model

The commands used to train the released models are provided in script train.sh. Here are some train tips:

• You should download CelebA to train WFEN. Please change the --dataroot to the path where your training images are stored.
• To train WFEN, we simply crop out faces from CelebA without pre-alignment, because for ultra low resolution face SR, it is difficult to pre-align the LR images.
• Please change the --name option for different experiments. Tensorboard records with the same name will be moved to check_points/log_archive, and the weight directory will only store weight history of latest experiment with the same name.
• If there's not enough memory, you can turn down the --batch_size.
• --gpus specify number of GPUs used to train. The script will use GPUs with more available memory first. To specify the GPU index, uncomment the export CUDA_VISIBLE_DEVICES=.

# Train Code
CUDA_VISIBLE_DEVICES=0,1 python train.py --gpus 2 --name wfen --model wfen \
    --Gnorm "bn" --lr 0.0002 --beta1 0.9 --scale_factor 8 --load_size 128 \
    --dataroot /path/to/datasets/CelebA --dataset_name celeba --batch_size 32 --total_epochs 150 \
    --visual_freq 100 --print_freq 10 --save_latest_freq 500

Acknowledgements

This code is built on Face-SPARNet. We thank the authors for sharing their codes.

:e-mail: Contact

If you have any question, please email lewj2408@gmail.com or cswjli@bupt.edu.cn