COM_COM_000166-Deepfake-Detection

Please find further details about MesoNet and YOLOv5 in the respective folders.

For train/val data please contact FaceForensics++ https://github.com/ondyari/FaceForensics#access

Process Flow

Data PreProcessing

Voting Mechanism

RESULTS

Accuracy_all5

Accuracy_individual

We have publised the reults of YOLOv5 on the here - http://kaldir.vc.in.tum.de/faceforensics_benchmark/

MesoNet Benchmarking

YOLOv5 Benchmarking

YOLOv5 with Voting Mechanism Benchmarking

Front End

In the YOLOv5 directory, one can simply execute python3 app.py to run flask server and launch site on localhost.