This is the implemention of Speech Adversarial Examples Attack with Generative Adversarial Networks
The pre-trained victim model is available for use and can be downloaded from here
The pre-trained target(generator) model is available for use and can be downloaded from here
Notice: The directory structure of the checkpoints file is as follows:
|-- checkpoints
|-- music_genres_generator
|-- target_blues
|-- target classical
|-- ...
|-- speech_common_generator
|-- target yes
|-- target no
|-- ...
|-- sampleCNN.pth
|-- wideResNet28_10.pth
To run:
python test_gan.py --data_dir original_speech.wav --target yes --checkpoint checkpoints