Inception-Score

• Inception score script for measuring quality of images generating from GAN

Method

• This repo derived from paper named Improved Techniques for Training GANs, In this paper, we can see more tricks and improved methods used for training GAN. Anyway, sometimes, they are helpful.

• Inception score is used to measure the quality of images generating from GAN model. It is considered as a good way to measure performance of GAN in quantity.

Requirements

• tensorflow 1.3.0

• python 2.7.12

• numpy 1.13.1

• scipy 0.17.0

Usages

download repo

$ git clone https://github.com/nnuyi/Inception-Score.git
$ cd Inception-Score

load data

• In this repo, I provide 30 images sampled from cifar10. All the images are stored in the data directory(named data).

• If you want to use your own data, images should be stored in data directory. And then launch the code.

launch repo

$ python inception_score.py

Results

• In this repo, you can easily get the result below:

  cifar10 images	cifar10 images
  inception score: mean:2.1532264 stddev:0.27045175	inception score: mean:2.1532264 stddev:0.27045175

References

• Code is derived from openai/improved-gan. Thanks all the way.

Contacts

• Email:computerscienceyyz@163.com