Team Name: dy/dx
Deepfakes are synthetic media in which a person in an existing image or video is replaced with someone else's likeness. This project leverages the XceptionNet architecture to detect such deepfakes with high accuracy.
- Problem Statement ID: 1683
- Problem Statement Title: Development of AI/ML based solution for detection of face-swap based deep fake videos
dydx_demo.mp4
This repository contains the implementation of the deepfake detection model using XceptionNet as presented in the paper: A. V and P. T. Joy, "Deepfake Detection Using XceptionNet," 2023 IEEE International Conference on Recent Advances in Systems Science and Engineering (RASSE), Kerala, India, 2023, pp. 1-5. Read the paper
-
Browser Extension: Develop an extension to detect deepfake content directly in the browser and provide user feedback.
-
Social Media Links Support: Integrate functionality to analyze and detect deepfakes in videos shared via social media links.
-
Detection Using Audio Anomalies: Add audio analysis to identify anomalies that may indicate deepfake content.
-
Frame/Frames Responsible for DeepFake Flag: Implement a feature to identify specific frames responsible for deepfake detection, enhancing the granularity of analysis.
You can download the trained models from the following link:
Clone this repository:
git clone https://github.com/dhruuvd-1704/dy-dx.git
cd dy-dx
Create a Terminal for Backend and run:
uvicorn backend:app --reload
Create a new Terminal for Frontend and run:
streamlit run app.py
WEB DEVELOPMENT
- Streamlit: For building interactive web interfaces for the deepfake detection system and blockchain integration.
- Web3: To interact with the Ethereum blockchain for querying and updating smart contracts.
BACKEND
- FastAPI: For handling video uploads, deepfake detection, and API responses efficiently.
Dhruv Desai
Atharva Humane
Niranjan More
Mithilesh Singh
Vaishnavi Hud
Kasturi Pawar