ethanhe42/gan-mnist

Generative Adversarial Network for MNIST with tensorflow

GAN on MNIST with TensorFlow

GitHub - yihui-he/GAN-MNIST: Generative Adversarial Network for MNIST with tensorflow

Tensorflow implementation

• All the codes in this project are mere replication of Theano version

Code

• Under face/ and mnist/
• model.py
• Definition of DCGAN model
• train.py
• Training the DCGAN model (and Generating samples time to time)
• util.py
• Image related utils

Dataset

• MNIST
• http://yann.lecun.com/exdb/mnist/
• CelebA Face dataset
• http://mmlab.ie.cuhk.edu.hk/projects/CelebA.html
• Download “img_align_celeba” images
• Set “face_image_path” in train.py according to the path of downloaded dataset

references

https://github.com/carpedm20/DCGAN-tensorflow

Citation

If you find the code useful in your research, please consider citing:

@InProceedings{He_2017_ICCV,
author = {He, Yihui and Zhang, Xiangyu and Sun, Jian},
title = {Channel Pruning for Accelerating Very Deep Neural Networks},
booktitle = {The IEEE International Conference on Computer Vision (ICCV)},
month = {Oct},
year = {2017}
}

gan-mnist's Issues

deconv2d

tf.nn.deconv2d is renamed as conv2d_transpose from tensorflow version 0.7

The relationship of the references to the code

What's the relationship between this code repo and the paper suggested for citation?

clipping raw value of discriminator

Hey, just browsed your code in mnist/model.py and I guess that you should not clip the output of the discriminator to 1e-7 and 1. - 1e-7.

In line 84 / 85 you pass the output of the descriminator (which is the result of conv2d_transpose) to your function bce, which clips the value. I guess you intended to clip the sigmoid.

Problem with load.py

I apologize if its a basic question, I'm new to Tensorflow. When I run train.py I get the following error while loading the MNIST dataset:

Traceback (most recent call last):
File "/home/m132/Desktop/Python/Learning GANs/GAN-MNIST-DC/train.py", line 19, in
trX, vaX, teX, trY, vaY, teY = mnist_with_valid_set()
File "/home/m132/Desktop/Python/Learning GANs/GAN-MNIST-DC/Load_Data.py", line 31, in mnist_with_valid_set
trX, teX, trY, teY = mnist()
File "/home/m132/Desktop/Python/Learning GANs/GAN-MNIST-DC/Load_Data.py", line 11, in mnist
trX = loaded[16:].reshape((60000,28*28)).astype(float)
ValueError: cannot reshape array of size 9912406 into shape (60000,784)

Any help would be appreciated.

Which tensorflow this code use

Could you tell me which version tensorflow used? I run "face/train.py", and got an error:

if v.op.type == "VarHandleOp":
AttributeError: 'filter' object has no attribute 'op'

I have used tensorflow 1.8.0 from pip in windows 10 1803.