Skip to content
Permalink
Branch: master
Commits on Jul 31, 2018
  1. Merge branch 'dev'

    ironhead
    ironhead committed Jul 31, 2018
  2. Merge branch 'features/update_readme' into dev

    ironhead
    ironhead committed Jul 31, 2018
  3. - update README (add two more GANs)

    ironhead
    ironhead committed Jul 31, 2018
Commits on Oct 8, 2017
  1. Merge branch 'dev'

    imironhead committed Oct 8, 2017
    * dev:
      - add readme for cyclegan
      - add methods to translate all images under specified directory
      - add methods for translation
      - change num_outputs of the first layer in generator to 64
      - add more flags.
      - workable solution
      - initial implementation
      - fix indent
      - add README for pix2pix
      - add detailed comments from the paper
      - refine summary creation method
      - fix: incorrect activation_fn and normalizer for the tail layer of the generator
      - add method for pix2pix translation (not training)
      - pix2pix initial commit, can not reproduce results that as good as paper's
  2. Merge branch 'features/cyclegan' into dev

    imironhead committed Oct 8, 2017
    * features/cyclegan:
      - add readme for cyclegan
      - add methods to translate all images under specified directory
      - add methods for translation
      - change num_outputs of the first layer in generator to 64
      - add more flags.
      - workable solution
      - initial implementation
  3. - add readme for cyclegan

    imironhead committed Oct 8, 2017
Commits on Oct 4, 2017
Commits on Oct 3, 2017
  1. - add more flags.

    imironhead committed Oct 3, 2017
Commits on Oct 2, 2017
  1. - workable solution

    imironhead committed Oct 2, 2017
Commits on Aug 25, 2017
  1. - initial implementation

    imironhead committed Aug 25, 2017
Commits on Aug 7, 2017
  1. Merge branch 'hotfixes/pix2pix' into dev

    imironhead committed Aug 7, 2017
    * hotfixes/pix2pix:
      - fix indent
  2. - fix indent

    imironhead committed Aug 7, 2017
Commits on Aug 6, 2017
  1. Merge branch 'features/pix2pix' into dev

    imironhead committed Aug 6, 2017
    * features/pix2pix:
      - add README for pix2pix
      - add detailed comments from the paper
      - refine summary creation method
      - fix: incorrect activation_fn and normalizer for the tail layer of the generator
      - add method for pix2pix translation (not training)
      - pix2pix initial commit, can not reproduce results that as good as paper's
  2. - add README for pix2pix

    imironhead committed Aug 6, 2017
Commits on Aug 5, 2017
Commits on Aug 4, 2017
Commits on Aug 3, 2017
Commits on Jun 14, 2017
  1. Merge branch 'dev'

    imironhead committed Jun 14, 2017
    * dev:
      - random crop image for began (on celeba dataset to increase training data) - refactor - ignore folder 'ckpts'
      - refactor (collect and group trainable variables)
      - refine flags, make learning rate configurable
      - add histogram summary of variables & gradients to dcgan
  2. Merge branch 'features/began_lr_decay' into dev

    imironhead committed Jun 14, 2017
    * features/began_lr_decay:
      - random crop image for began (on celeba dataset to increase training data) - refactor - ignore folder 'ckpts'
  3. - random crop image for began (on celeba dataset to increase training…

    imironhead committed Jun 14, 2017
    … data)
    
    - refactor
    - ignore folder 'ckpts'
Commits on May 29, 2017
  1. Merge branch 'features/dcgan_gradients' into dev

    imironhead committed May 29, 2017
    * features/dcgan_gradients:
      - refactor (collect and group trainable variables)
      - refine flags, make learning rate configurable
      - add histogram summary of variables & gradients to dcgan
Commits on May 20, 2017
  1. Merge branch 'hotfixes/typo'

    imironhead committed May 20, 2017
    * hotfixes/typo:
      - typo, highlight methods of md is not the same with blog one.
  2. Merge branch 'hotfixes/typo' into dev

    imironhead committed May 20, 2017
    * hotfixes/typo:
      - typo, highlight methods of md is not the same with blog one.
      - fix typo
  3. Merge branch 'hotfixes/typo'

    imironhead committed May 20, 2017
    * hotfixes/typo:
      - fix typo
  4. - fix typo

    imironhead committed May 20, 2017
  5. Merge branch 'dev'

    imironhead committed May 20, 2017
    * dev:
      - add/update README
      - refactor shift.py. 3 modes are implemented: -- interp_linear: generate an image stride linear interpolated with 2 vectors. -- interp_bilinear: generate an image grid bilinear interpolated with 4 vectors. -- random: generate a random image grid
      - refactor
      - use l1 loss to find corresponding z of faces
      - add 'random' experiment (bilinear interpolate 4 face vectors)
      - add shift.py to do interpolations.
      - add embed.py to map a face to its embedding vector (by optimazation)
      - split began.py to began.py and train.py. keep began model in began.py.
      - refactor, add more summaries (autoencoded real/fake images, convergence measure).
      - refactor
      - read training set path from parameters
      - add cropper to crop origin face images (so we do not have to crop and scale during training)
      - refactor a little bit
      - initial version of began (collapse happened)
      - reduce lambda to train on lsun bedroom training set. - save checkpoints every 500 steps.
  6. Merge branch 'features/began' into dev

    imironhead committed May 20, 2017
    * features/began:
      - add/update README
      - refactor shift.py. 3 modes are implemented: -- interp_linear: generate an image stride linear interpolated with 2 vectors. -- interp_bilinear: generate an image grid bilinear interpolated with 4 vectors. -- random: generate a random image grid
      - refactor
      - use l1 loss to find corresponding z of faces
      - add 'random' experiment (bilinear interpolate 4 face vectors)
      - add shift.py to do interpolations.
      - add embed.py to map a face to its embedding vector (by optimazation)
      - split began.py to began.py and train.py. keep began model in began.py.
      - refactor, add more summaries (autoencoded real/fake images, convergence measure).
      - refactor
      - read training set path from parameters
      - add cropper to crop origin face images (so we do not have to crop and scale during training)
      - refactor a little bit
      - initial version of began (collapse happened)
Older
You can’t perform that action at this time.