gouravsinghbais/deepfake-image-clssification

Identifying the GANs generated image.

Deepfake-Image-Classification

Identifying the GANs generated image.

Dependencies and Installation Guide

Programming Language:

• Python 3.6

Libraries needed to Run this project:

• tensorlow-1.13.0
• opencv-python
• keras-2.3.0
• matplotlib
• os
• mtcnn
• pillow

Required model architecture and weights files:

you can download the weights and model architecture file from here: https://drive.google.com/open?id=1KujeIcVzoRWHgBlO-BzpClxhujUbeDmV

Dataset:

Dataset for training the model are available here: https://drive.google.com/file/d/1smby8vBB0g8bNtUsQ10OdjF-4FK9k_GS/view

Instructions to train the model:

• Dataset which is provided in the link is already preprocessed and it contains images which has only faces i.e. MTCNN is already applied.
• You need to download the dataset and just run the model training section since face detection part is already done.
• give the path of folders into the ImageDataGenerator and run the model it will start training.

Insutuctions to test the model:

• You will to download the model.json and model.h5 file and will have to give the path of these files in the testing of the notebook.
• Then choose the image on which you want to make the prediction give that imaeg path in the testing section and predictions will be generated.

References:

• https://www.researchgate.net/publication/327905310_Forensics_Face_Detection_From_GANs_Using_Convolutional_Neural_Network?enrichId=rgreq-fa3b4b45646d1695236abfc1b3a44062-XXX&enrichSource=Y292ZXJQYWdlOzMyNzkwNTMxMDtBUzo2NzUzNzI5NjYzNTkwNTZAMTUzODAzMjg3MTc5MA==&el=1_x_3&_esc=publicationCoverPdf

You can read my blog on Fake Face Classification:

https://medium.com/@gouravbais08/fake-face-image-classification-91a6225e708d

<script src="https://gist.github.com/gouravsinghbais/656938bc2ae8818c10cc39982405fecd.js?file=gist.md"></script>