infogan-pytorch's Introduction

infoGAN-pytorch

Implementation of infoGAN using PyTorch

Results

Catagorical variable(K) and continue variable (c1)
Row represents categorical variable from K = 0 to K = 9 (top to buttom) to characterize digits.
Column represents continue variable (c1) (Width) varying from -1 to 1(left to right).

Catagorical variable(K) and continue variable (c2)
Row represents categorical variable from K = 0 to K = 9 (top to buttom) to characterize digits.
Column represents continue variable (c2) (Rotation) varying from -1 to 1(left to right).

infogan-pytorch's People

Contributors

Stargazers

Watchers

infogan-pytorch's Issues

what the log_gaussian loss's mean?

class log_gaussian:
def call(self, x, mu, var):
logli = -0.5*(var.mul(2*np.pi)+1e-6).log() -
(x-mu).pow(2).div(var.mul(2.0)+1e-6)
return logli.sum(1).mean().mul(-1)

1、What do mu and var represent?
2、I'm doubt about the log_gaussion's theory.

thank you!

Arbitrary size images

Thanks for this awesome simple to understand repo.

I'm trying to use this for data augmentation, for a new dataset with size 128x128, and a latent variable which is a vector of size 37.

I tried to play a bit with the Conv2d parameters, but unfortunately no configuration I tried made it run without errors. I was wondering if you can please add a size-dependent convolution to the source code, so in the future anyone with any size can just input their size and get this working correctly.

Thanks

Recommend Projects

pianomania / infogan-pytorch Goto Github PK

infogan-pytorch's Introduction

infoGAN-pytorch

Results

infogan-pytorch's People

Contributors

Stargazers

Watchers

Forkers

infogan-pytorch's Issues

what the log_gaussian loss's mean?

Arbitrary size images

Recommend Projects

React

Vue.js

Typescript

TensorFlow

Django

Laravel

D3

Recommend Topics

javascript

web

server

Machine learning

Visualization

Game

Recommend Org

Facebook

Microsoft

Google

Alibaba

D3

Tencent