同学，我是TensorFlow 1.8版本。之前用原作者的程序会报错，看你留言tf 1.6+能用。
现在运行程序显示

Downloading http://download.tensorflow.org/models/frozen_inception_v1_2015_12_05.tar.gz 63.3%
我是新手，不知道如何改程序才能应用到自己的36张256*256的测试图片上。谢谢

INCEPTION_URL = "http://download.tensorflow.org/models/frozen_inception_v1_2015_12_05.tar.gz"

tfgan.eval.get_graph_def_from_url_tarball(url=INCEPTION_URL,filename=INCEPTION_FROZEN_GRAPH,tar_filename=os.path.basename(INCEPTION_URL))

执行上述代码会出现 http error 409： conflict

Hi, i test CIFAR-10 50000 training images using your get_inception_score, but the results is only about 9(mean) for 1 split. Your mentioned you get the results over 11. I have check the input numpy is in the range of [0,255] and shape of [50000,3,32,32], i am not sure why. So could you provide the dataloader part to help my testing?

Hi @tsc2017,

I have gone through your code, and I find that you are not resizing the input images before calculating their activation scores from the InceptionNet.
I found your repo from this discussion at openai/improved-gan#29
For some reason, I am not getting correct IS value from this repo, so was looking for other codes and found this one.
Could you please let me know if the function requires resized images to size [229 x 229 x 3] is it okay to send it [32 x 32 x 3] CIFAR-10 images?

Best regards,
@akanimax

with your help of this code, I can put it into my model to evaluate my images generated by GANs, but the problem confused me for almost a week is the input of the code is array, but I wanna input of the function in my model is 4D tensor, and I also wanna put the result of the IS to the tensorboard, I've tried several ways like put the transference of the function in Session to transfer tensor to array as the parameters of the function (but the error is about the initialization). can u help me handle this problems ?