buriburisuri / ac-gan Goto Github PK

A tensorflow implementation of google's AC-GAN ( Auxiliary Classifier GAN ).

License: MIT License

Python 100.00%

ac-gan's Introduction

AC-GAN ( Auxiliary Classifier GAN )

A tensorflow implementation of Augustus Odena (at Google Brains) et al's "Conditional Image Synthesis With Auxiliary Classifier GANs" paper ) I've already implemented this kind of GAN structure last Sep. (See : Supervised InfoGAN tensorflow implementation) I said that I had added supervised loss(in this paper auxiliary classifier) to InfoGAN structure to achieve the consistency of generated categories and training stability. And the result was promising. It helped the model to generate consistent categories and to converge fast compared to InfoGAN. The paper which propose the same architecture by researchers at Google Brains was published last Oct.
They've provided many results from diverse data set and tried to generate big images and explained why this architecture works plausibly. I think this paper is awesome. But they did not provide source codes and I re-introduce this codes.

AC-GAN architecture( from the paper. )

Dependencies

tensorflow >= rc0.11
sugartensor >= 0.0.1.7

Training the network

Execute


python train.py

to train the network. You can see the result ckpt files and log files in the 'asset/train' directory. Launch tensorboard --logdir asset/train/log to monitor training process.

Generating image

Execute


python generate.py

to generate sample image. The 'sample.png' file will be generated in the 'asset/train' directory.

Generated image sample

This image was generated by AC-GAN network.

And this image was generated with categorical auxiliary classifier.

And this image was generated by continuous auxiliary classifier. You can see the rotation change along the X axis and the thickness change along the Y axis.

The following image is the loss chart in the training process. This looks more stable than my original GAN implementation.

Other resources

Authors

Namju Kim ([email protected]) at Jamonglabs Co., Ltd.

ac-gan's People

Contributors

Stargazers

Watchers

Forkers

shaoyizhang wanjinchang ahn19 benjamesbabala fducau ml-lab vyraun amoliu federicov simmoncn 309972460 kgl-prml recluze zhangyuancv kli-casia chunniunai220ml wenxuanliu karimpedia xiangyongcao gitsamshi scott-vsi yfliao jiajiadf shouguicai jsliuqun beijinggao cllarence zilongzhong gongqingyi-github hamedmp mxl1990 kjeanclaude rremani richardkelley sjtucsly henryslzhao hdnse4798 mengmengbai xhchrn shuolongbj zhl10154 kgzy libardo1 sunxingxingtf gongcaiwang vonzunlei drzhanying xychen9459 jacoblee121 csu-gh leochencipher iqbal-chowdhury sungjunyoung jkkorea yangxue0827 rahulptel alexleethinker kangxz praveenkumarchandaliya fengxhao sesebuckin toxato startime-h rxia jianfeidahai shubhampachori12110095 pierluigidovesi zqcchris deep0learning isr-wang hbcbh1999 afcarl wangyida hhsummerwind ruiatelsevier duyamin sibangde zlpsls q145492675 namapolo ai-gzhu ligangzheng devashishjoshi pingcsu meiyangzhang danxiangjie zccdl mgong2 srinivasgutta7 pwichmann phoebexy etiennexiong anas-alamri elyesmanai lslsbj89 duke24k gky15 pkulwj1994 iq-scm

ac-gan's Issues

Training problem!

Thank for sharing this code, but when I trying to train this model with python train.py, the following error occurs:

Traceback (most recent call last):
File "train.py", line 86, in
y_disc = tf.concat(0, [y, y * 0])
File "/home/hao/anaconda2/lib/python2.7/site-packages/tensorflow/python/ops/array_ops.py", line 1062, in concat
).assert_is_compatible_with(tensor_shape.scalar())
File "/home/hao/anaconda2/lib/python2.7/site-packages/tensorflow/python/framework/tensor_shape.py", line 737, in assert_is_compatible_with
raise ValueError("Shapes %s and %s are incompatible" % (self, other))
ValueError: Shapes (2, 32) and () are incompatible

generate problem

Traceback (most recent call last):
File "generate.py", line 98, in
fig_name='fake.png')
File "generate.py", line 73, in run_generator
saver.restore(sess, tf.train.latest_checkpoint('asset/train'))
File "/home/xujian/.local/lib/python2.7/site-packages/tensorflow/python/training/saver.py", line 1439, in restore
{self.saver_def.filename_tensor_name: save_path})
File "/home/xujian/.local/lib/python2.7/site-packages/tensorflow/python/client/session.py", line 767, in run
run_metadata_ptr)
File "/home/xujian/.local/lib/python2.7/site-packages/tensorflow/python/client/session.py", line 965, in _run
feed_dict_string, options, run_metadata)
File "/home/xujian/.local/lib/python2.7/site-packages/tensorflow/python/client/session.py", line 1015, in _do_run
target_list, options, run_metadata)
File "/home/xujian/.local/lib/python2.7/site-packages/tensorflow/python/client/session.py", line 1035, in _do_call
raise type(e)(node_def, op, message)
tensorflow.python.framework.errors_impl.InternalError: Unable to get element from the feed as bytes.

good work!

thanks

data problem

Extracting ./asset/data/mnist/train-images-idx3-ubyte.gz
Traceback (most recent call last):
File "train.py", line 75, in
data = tf.sg_data.Mnist(batch_size=batch_size)
File "/home/xujian/ac-gan-master/ac-gan-master/sugartensor/sg_data.py", line 40, in init
data_set = input_data.read_data_sets(Mnist._data_dir, reshape=reshape, one_hot=one_hot)
File "/home/xujian/.local/lib/python2.7/site-packages/tensorflow/contrib/learn/python/learn/datasets/mnist.py", line 213, in read_data_sets
train_images = extract_images(f)
File "/home/xujian/.local/lib/python2.7/site-packages/tensorflow/contrib/learn/python/learn/datasets/mnist.py", line 53, in extract_images
magic = _read32(bytestream)
File "/home/xujian/.local/lib/python2.7/site-packages/tensorflow/contrib/learn/python/learn/datasets/mnist.py", line 35, in _read32
return numpy.frombuffer(bytestream.read(4), dtype=dt)[0]
File "/usr/lib/python2.7/gzip.py", line 268, in read
self._read(readsize)
File "/usr/lib/python2.7/gzip.py", line 303, in _read
self._read_gzip_header()
File "/usr/lib/python2.7/gzip.py", line 197, in _read_gzip_header
raise IOError, 'Not a gzipped file'
IOError: Not a gzipped file

Loss to update G net

As is known, only generated samples are used to calculate adversarial loss for updating G net in classic gan. Here, I see that real image data was also used to calculate classification loss to update G net? Is it ok to do this?

which paper contains the architecture of these gans?

The picture in Readme contains conditional gan, semi-supervised gan, info gan and acgan. But it seems that the paper of acgan does not have this picture. So which paper contains this picture?

Some question about the code

Hi @buriburisuri , your code ac-gan loss, Here the loss which need to be optimized seems not be a 0d tensor, didn't it need to sg_mean() with the loss_c_r and loss_c_d?

An question about the code

the "y_disc = tf.concat(0, [y, y * 0])" in 82 lines seems unused?

2.You divide the class info into categorical latent code and conditional latent code, Is this thought from InfoGan? I don't find the paper do it.