**
Deepfake detection by using Xception Network architecture was proposed by FaceForensic++ (https://arxiv.org/pdf/1901.08971.pdf). Our implementation is also based on the parameters and technique described in the FaceForensic++ paper. It is also considered as a benchmark to evaluate the performance of new detection techniques.
** Dataset:**
The processed dataset could be downloaded from this link: https://seafile.idmt.fraunhofer.de/u/d/b639276d15b9423bb11f/
**
Step 1
- Extract the frames from the videos using the file
"Rename and Frame Extraction.ipynb"by only adding the path to the directory.
Step 2:
- Follow the instructions in file
"Data_Analysis.ipynb"to create the test, train, and cv directory with only the cropped faces from the frames extracted earlier. - Also, follow further instruction to label and store the data for further processing.
Step 3:
- Follow instructions in file
"Model Training.ipynb"to train the model for deepfake detection.
Step 4:
- Follow the instructions in file
"Test_Xception.ipynb"to evaluate the model's performance on the unseen video data.