Fig. 3From: Towards the transferable audio adversarial attack via ensemble methodsTransferable attacks on speech recognition models. Yellow icons mark benign samples and red icons mark adversarial samplesBack to article page