Scribd
Upload
145 views19 pages

ESE Presentation

This document describes a deep learning model to classify videos as deepfakes or authentic. It discusses deepfakes, the need for detection, how deepfakes are created, the proposed system architecture including data preprocessing, model architecture using ResNext-50 and LSTM, training and prediction workflows, tools used like PyTorch and Google Cloud, and results on two datasets showing over 90% accuracy with more frames.

Uploaded by

ketan itcell
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PPTX, PDF, TXT or read online on Scribd
145 views19 pages

ESE Presentation

This document describes a deep learning model to classify videos as deepfakes or authentic. It discusses deepfakes, the need for detection, how deepfakes are created, the proposed system architecture including data preprocessing, model architecture using ResNext-50 and LSTM, training and prediction workflows, tools used like PyTorch and Google Cloud, and results on two datasets showing over 90% accuracy with more frames.

Uploaded by

ketan itcell
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PPTX, PDF, TXT or read online on Scribd
You are on page 1/ 19

By

Abhijit Jadhav
Abhishek Patange
Jay Patel
Hitendra Pail

Guided by:
Ms. Manjushri Mahajan
Problem Statement

To Design and Develop a


Deep Learning algorithm
to classify the video as
deepfake or pristine.
Introduction Training Workflow
01 06

Prediction Workflow
System Architecture 07
02

Dataset Exploration 08 Tools and Technologies


03

Pre-processing 09 Results
04

05 Model Architecture 10 Demo


 Introduction

• Deep fake is a technique for


human image synthesis based on
artificial intelligence.
• Deep fakes are created by
combing and superimposing
existing images and videos onto
source images or videos using a
deep learning technique known
as generative adversarial
network.
Can we detect Deep fakes with naked eyes?
• Why Deep Fake Detection ?

• Fake News
• Malicious hoaxes
• Financial fraud
• Celebrity unusual video
• Revenge porn
• Politician videos
 How Deep Fakes Are Created ?

Tools for deep fake creation.


• Faceswap
• Faceit
• DeepFaceLab
• DeepfakeCapsuleGAN
• Large resolution
facemasked
System
Architecture
Data-set
Exploration
SPLIT VIDEO
1 INTO FRAMES

2 FACE
DETECTION

Pre-
3 CROPING FACE
processing
CREATING NEW
4 FACE CROPPED
VIDEO

SAVING THE
5 FACE CROPPED
VIDEO
Model Architecture

ResNext-50 1 LSTM layer with


2048 shape input
vector and 2048
latent features along
with 0.4
chance of dropout

Sequential Layer
and ReLU
Activation function
Training Workflow
Prediction
Workflow
Programming Languages
• Python3
• JavaScript

Programming Frameworks
• PyTorch
• Django

Tools and IDE


• Google
Technologies • Jupyter Notebook
• Visual Studio Code

Cloud Services
• Google Cloud Platform

Version Control
• Git
No of Sequence
Model Name Dataset Accuracy
Videos Length
model_90_acc_20_frames_FF_data 20 90.95477387

model_95_acc_40_frames_FF_data 40 95.22613065

model_97_acc_60_frames_FF_data FaceForensic++ 60 97.48743719


2000

model_97_acc_80_frames_FF_data 80 97.73366834

model_97_acc_100_frames_FF_data 100 97.76180905


Results model_84_acc_10_frames_final_data 10 84. 662519

model_87_acc_20_frames_final_data 20 87.79160186

model_89_acc_40_frames_final_data 40 89.3468118195956
Our Dataset 6000
model_91_acc_60_frames_final_data 60 91.5909797822706

model_92_acc_80_frames_final_data 80 92.4981855883877

model_93_acc_100_frames_final_data 92.10883877
100

You might also like