Ensembling of two reshaped and tested models for DeepFake detection based on [1-2], while combining visual artifacts and interpolated/compressed frequency features. Final model tested on the UADFV inter-database.

Authors: Sebastián Cajas, Julián Salazar, Zhanbota Bissaliyeva

References:

[1] F. Matern, Ch.Riess, M.Stamminger, “Exploiting Visual Artifacts to Expose Deepfakes and Face Manipulations”, Friedrich-Alexander University Erlangen-Nuremberg, 2019

[2] Durall, R., Keuper, M., Pfreundt, F. J., & Keuper, J. (2019). Unmasking deepfakes with simple features. arXiv preprint arXiv:1911.00686.