Welcome to NeuroForge, an advanced AI-powered platform designed to detect deepfakes in videos. NeuroForge leverages state-of-the-art deep learning models to analyze video content and determine its authenticity, offering explanations for detected inconsistencies and manipulation techniques.
NeuroForge is a user-friendly, intuitive deepfake detection tool. This platform enables users to upload videos, receive real-time analysis results, and understand why the content might be classified as real or fake. It is designed for a range of users, from media outlets and content creators to security organizations, aiming to preserve the authenticity of visual media in an age of increasing synthetic content.
- NeuroForge offers an intuitive interface for video upload and results display, including detailed detection reports with confidence scores.
- Efficiently analyzes video content, providing real-time detection results with insightful explanations about potential manipulation techniques.
- Provides detailed explanations for each classification, highlighting inconsistencies in lighting, motion, and facial features to help users understand deepfake characteristics.
- Ideal for enterprise clients, supporting batch video processing to verify content authenticity at scale.
- Each analysis includes a confidence score indicating the likelihood of the content being real or fake, alongside a polarity analysis for a comprehensive evaluation.
- Committed to ethical AI and data privacy, NeuroForge follows industry standards for handling and analyzing video content securely.
To run NeuroForge, make sure you have the following dependencies installed:
Python >= 3.8StreamlitTensorFlowOpenCVNumpy
- Clone this repository:
git clone hhttps://github.com/ParivalavanIT/deepfake-detection.git
- Change directory to the project folder:
cd deepfake-detection - Install required packages:
pip install -r requirements.txt
- Run the Streamlit application:
streamlit run app.py
- Open your browser to view the app at
http://localhost:8501. - Upload a video file (supported formats: mp4, avi, mov, mkv, mpeg) and start the analysis.
- The app will display a real-time confidence score, classification as "Real" or "Fake," and detailed explanations if inconsistencies are detected.
streamlittensorflowopencv-pythonnumpyrequests(for fetching animations or external resources)
- FaceNet embeddings: Used for extracting facial features.
- LSTM-based Sequence Model: For analyzing temporal information and detecting anomalies within video sequences.
- InceptionResNetV2: Serves as the base model for embedding extraction with dimensions
(160, 160, 3)for frame inputs.
- The app provides detailed explanation-based feedback, outlining detected inconsistencies in facial features, lighting, or motion patterns. Each result is accompanied by confidence scores and suggestions to ensure users receive a well-rounded evaluation.
- Media Outlets: Quickly verify video authenticity to prevent misinformation.
- Security and Surveillance: Identify deepfakes in surveillance footage for threat detection.
- Legal Evidence Verification: Confirm credibility of video evidence for legal applications.
NeuroForge represents a pioneering approach to deepfake detection, providing real-time feedback and explanation-driven insights. This platform is an invaluable tool in today's digital age, assisting individuals and organizations in preserving the integrity of video content and promoting media authenticity.
For questions or support, reach out at [parivalavan2345@gmail.com].