/Stacked-GAN-Face-SR

Stacked GAN face super resolution

Primary LanguagePythonMIT LicenseMIT

Stacked-GAN-Face-SR

Face Synthesis and Super Resolution Using Stacked Generative Adversarial Network

1. Dataset generate

  • a.Using HD CelebA Cropper to generate the dataset raw data.
  • b.Use the default 0.7 face_factor.
  • c.Copy the raw data to dataset folder.
  • c.Run 1.get_celeb_train.py to generate training, validating and testing data

2. Training

  • a.Run the training command python 2.train_and_test.py --phase train --dataset_name yourData --origin_size 64 --image_size 256
  • b.Check the sample folder during training to make sure the result is correct.

3. Testing

  • a.Run the testing command python 2.train_and_test.py --phase test --dataset_name yourData --origin_size 64 --image_size 256
  • b.Check the generated test folder for the results

4. Calculate FID

  • a.The FID calculating code is from here
  • b.Run 3.split_testset.py to seperate blur tesing images and ground truth
  • c.Run the FID calculating command python 4.fid_score.py dataset/yourData/gt test

5. FID results

Table 1 FID score compare in different resolution transformation

name 16 to 64 32 to 128 64 to 256
Wavelet SRNet 45.952 54.533 31.929
Cycle GAN 63.837 32.228 42.218
SRN-Deblur 125.576 41.463 12.729
ours 51.646 29.685 13.431

6. Images results

Fig