This project page describes our paper at the International Conference on Machine Learning (ICML), Synthetic Realities: Deep Learning for Detecting AudioVisual Fakes Workshop.

David Güera	Sriram Baireddy	Paolo Bestagini	Stefano Turbaro	Edward J. Delp

We propose to identify forged videos by analyzing their multimedia stream descriptors with simple binary classifiers, completely avoiding the pixel space.

This repo details all the processes of our proposed pipeline to train models, test them, and evaluate them.

Download our paper at the International Conference on Machine Learning (ICML), Synthetic Realities: Deep Learning for Detecting AudioVisual Fakes Workshop by clicking here. Please cite it with the following bibtex code:

@Article{Guera2019_ICMLW,
    author   = {D. G\"{u}era and S. Baireddy and P. Bestagini and S. Tubaro and E. J. Delp},
    journal  = {International Conference on Machine Learning (ICML), Synthetic Realities: Deep Learning for Detecting AudioVisual Fakes Workshop},
    title    = {We Need No Pixels: Video Manipulation Detection Using Stream Descriptors},
    year     = {2019},
    month    = {June},
    note     = {{Long Beach, CA}}
}

You may also want to refer to our publication with the more human-friendly Chicago style:

David Güera, Sriram Baireddy, Paolo Bestagini, Stefano Turbaro, and Edward J. Delp. "We Need No Pixels: Video Manipulation Detection Using Stream Descriptors." In Proceedings of the International Conference on Machine Learning (ICML), Synthetic Realities: Deep Learning for Detecting AudioVisual Fakes Workshop. 2019.

If you have any general question about our work or code which may be of interest to other researchers, please use the issues section on this git repo. You can also send us an e-mail at [email protected].