直接运行代码进行测试无法得到较好的效果 #43
Description
作者您好！我自己运行得到的效果很差，比如说我选择了inception model，得到的效果如下所示。我想请问是我进行测试的图片数量太少导致这样的结果还是一些其他原因。
*Transfer to mixer-b
Accuracy on benign examples: 70.0%
Accuracy on adversarial examples: 30.0%
*Transfer to mixer-l
Accuracy on benign examples: 70.0%
Accuracy on adversarial examples: 10.0%
*Transfer to tf2torch_adv_inception_v3
Accuracy on benign examples: 70.0%
Accuracy on adversarial examples: 10.0%
*Transfer to tf2torch_ens3_adv_inc_v3
Accuracy on benign examples: 90.0%
Accuracy on adversarial examples: 0.0%
*Transfer to tf2torch_ens4_adv_inc_v3
Accuracy on benign examples: 70.0%
Accuracy on adversarial examples: 10.0%
*Transfer to tf2torch_ens_adv_inc_res_v2
Accuracy on benign examples: 100.0%
Accuracy on adversarial examples: 20.0%
clean_accuracy: 100.0 90.0 80.0 100.0 100.0 90.0 100.0 90.0 100.0 70.0 70.0 70.0 90.0 70.0 100.0
adv_accuracy: 30.0 30.0 30.0 10.0 50.0 30.0 40.0 40.0 30.0 30.0 10.0 10.0 0.0 10.0 20.0
*fid: 287.84357605437066
期待您的答复，祝您生活愉快！