Merle314 / Supervised-Pixel-Wise-GAN

Supervised Pixel-Wise GAN for Face Super Resolution in PyTorch

SPGAN-PyTorch

Project | Arxiv | PyTorch

PyTorch implementation of Supervised Pixel-Wise GAN for Face Super Resolution.

Prerequisites

• Python 3.5
• PyTorch 0.4

Datasets

• VGGFace2
• CelebA
• LFW
• Helen

Run

Use the default hyparameters except changing the parameter "upscale" according to the expected upscaling factor(2, 3, 4 for 4, 8, 16 upcaling factors, respectively).

python main.py --ngpu=1 --test --start_epoch=0 --test_iter=1000 --batchSize=64 --test_batchSize=32 --nrow=4 --upscale=3 --input_height=128 --output_height=128 --crop_height=128 --lr=2e-4 --nEpochs=500 --cuda

Result

Comparison with state-of-the-art methods through 8x super-resolution 16x16 input faces. Zooming-in details of important face parts confirm the effectiveness of our SPGAN.

---

---

Comparison with state-of-the-art methods through 16x super-resolution 8x8 input faces. Please zooming-in for details.

Citation

If you use our codes, please cite the following paper:

Acknowledgments

Code borrows heavily from WaveeltSRNet.