Interpreting Adversarially Trained Convolutional Neural Networks

個人總結: 其實說白了就是: AT-CNNs are more biased towards global structures, such as shapes and edges。        AT-CNN對圖像的整體結構,比如形狀,邊緣更加敏感,可以更好的捕捉到這些信息。 AT-CNNs are less sensitive to the texture distortion and foc
相關文章
相關標籤/搜索