ylhz / tf_to_pytorch_model Goto Github PK

Model Accuracy in readme raises an issue of misleading.

For example, [tf_inception_v3] means that this model is converted from tf. But I just think that It is the tf models.

作者，你好 ，batchsize = 1的时候，传入模型获得logits的有的模型维度为[1,1001] 有的模型维度为[1001] 如何修改呢？

Please can you tell are the images here the sames used in NIPS 2017 adversarial competition? Thank you.

Thank you for sharing！
when source model is resnet50 or resnet101 or resnet152, the attack success rate( ASR) is has dropped significantly in the case of white-box attack.
for example, when adversarial samples are generated by Resnet50, as shown in the figure，the ASR for resnet50 is 54.2%.

Hi, I have been borrowing your framework to test my method recently, and I want to ask if these models support other sizes such as 224x244 input size, and I have encountered an error in this part

Could u add VGG19 model to your list? thx!

The super of DataLoader should be torch.utils.data.Dataset rather than DataLoader

您好 ，我想问一下转换后的模型怎么下载呀。我下载npy文件显示不能下载无法授权。请问可否分享我一份呢？我的邮箱是[email protected]谢谢