This repository contains samples of audio files used in:
Cross-Representation Transferability of Adversarial Attacks: From Spectrograms to Audio Waveforms
Visit https://karlmiko.netlify.app/papers/cross_representation/ to hear the samples!
Follow the notebooks, from 0 to 6. The dataset (GTZan) is not included in the repository so you must download it from http://marsyas.info/downloads/datasets.html.
To listen to the reconstructed audio, open the listen notebook.
The load_test notebook is for testing the performances of the CNNs.
If any questions arise do not hesitate in contacting me at [email protected].