testing use different kind image, but results are the same 'cat' about googlenet-inception

[duyueshengmx]

Additional infos: I have recoveried the test photos according to you dataset of 'cifar-10-python.tar.gz', and I found the test photos are too smaller, it's hard to Identified by my eyes, So maybe this is the reason of all my predict results are 'cat'. So I want to know, how about your predict results when use your pre-trained model?
Could you please give some help? Thank you very much!

[conan7882]

You may use the model trained on ImageNet, which can be found here:https://github.com/conan7882/GoogLeNet-Inception#imagenet-classification