README.md

metadata

title: FactSight
emoji: 🚀
colorFrom: blue
colorTo: purple
sdk: docker

FactSight: Advanced Fake News Detection System

📋 Table of Contents

• Overview
• Architecture
• Installation
• Usage
• Services & Models Deep Dive
  • Text Analysis Services
  • Image Analysis Services
  • Fact-Checking Integration
• Model Performance
• Demo Media Files
• Configuration
• API Endpoints
• Contributing
• License

🎯 Overview

FactSight is a comprehensive fake news detection system that combines multiple AI/ML technologies to analyze news content for authenticity. The system performs multi-modal analysis using both text and image content to determine if news articles are genuine or fabricated.

Key Features

• 🔍 Multi-Modal Analysis: Combines text and image analysis for comprehensive fact-checking
• 🤖 AI Content Detection: Identifies AI-generated text and images
• 🧠 Deep Learning Models: Uses state-of-the-art neural networks for classification
• 🔗 External Fact-Checking: Integrates with Google Fact Check Tools API
• 📊 Emotion Analysis: Detects sensationalism in news content
• 👤 Face Analysis: Analyzes faces in images for deepfake detection
• 🌐 Web Interface: User-friendly web application for easy analysis

🏗️ Architecture

┌─────────────────────────────────────────────────────────────────┐
│                        FactSight System                         │
├─────────────────────────────────────────────────────────────────┤
│  ┌─────────────┐  ┌─────────────┐  ┌─────────────┐              │
│  │   Flask     │  │   Core      │  │  Services   │              │
│  │   Web App   │  │  Manager    │  │ & Models    │              │
│  └─────────────┘  └─────────────┘  └─────────────┘              │
├─────────────────────────────────────────────────────────────────┤
│  ┌─────────────┐  ┌─────────────┐  ┌─────────────┐              │
│  │   Text      │  │   Image     │  │   Fact      │              │
│  │  Analysis   │  │  Analysis   │  │  Checking   │              │
│  └─────────────┘  └─────────────┘  └─────────────┘              │
├─────────────────────────────────────────────────────────────────┤
│  ┌─────────────┐  ┌─────────────┐  ┌─────────────┐              │
│  │   LSTM      │  │ EfficientNet│  │   Google    │              │
│  │ Classifier  │  │    B3       │  │ Fact Check  │              │
│  └─────────────┘  └─────────────┘  └─────────────┘              │
└─────────────────────────────────────────────────────────────────┘

🚀 Installation

Prerequisites

• Python 3.11+
• CUDA-compatible GPU (recommended for faster inference)
• 8GB+ RAM
• 4GB+ free disk space

Setup Instructions

1. Clone the repository

   git clone <repository-url>
   cd FactSight
   

2. Install dependencies

   pip install -r requirements.txt
   

3. Download required models from HUGGINGFACE (if not included)

   # Models will be automatically downloaded on first run
   # or placed in the models/ directory
   

4. Configure environment

   # Set your Google Fact Check API key in config.py
   FACT_API_KEY = "your-api-key-here"
   

5. Run the application

   python app.py
   

6. Access the web interface

   Open http://localhost:5000 in your browser
   

📖 Usage

Web Interface

1. Navigate to the homepage (/)
2. Paste news text in the text area (max 10,000 characters)
3. Upload images related to the news (drag & drop or browse)
4. Click "Analyze" to start the fact-checking process
5. View detailed results including:
   • Overall authenticity score
   • Text analysis breakdown
   • Image analysis results
   • Fact-checking verification
   • Emotion analysis

API Usage

import requests

# Submit for analysis
response = requests.post('http://localhost:5000/analyze', json={
    'text': 'Your news article text here...',
    'images': [
        {'data': 'base64-encoded-image-data'}
    ]
})

analysis_id = response.json()['analysis_id']

# Get results
results = requests.get(f'http://localhost:5000/analysis/{analysis_id}')

🔬 Services & Models Deep Dive

Text Analysis Services

1. AI Text Detection (ai_text_service.py)

• Model: Custom-trained classifier [placeholder for url]
• Architecture: Joblib-based scikit-learn model
• Performance: 99% F1-Score
• Purpose: Distinguishes between AI-generated and human-written text
• Features: Text preprocessing, feature extraction, binary classification

2. Fake News Classification (fake_text_news_service.py)

• Model: Custom LSTM Neural Network [placeholder for url]
• Architecture:
  • Embedding layer (100 dimensions)
  • Bidirectional LSTM (128 hidden units)
  • Dropout regularization (0.5)
  • Sigmoid output layer
• Performance: 99% F1-Score
• Purpose: Classifies news as fake or real
• Features:
  • Text cleaning and preprocessing
  • Contraction expansion
  • Stopword removal
  • Sequence padding/truncation (300 tokens)

3. Emotion Analysis (text_emotion_service.py)

• Model: j-hartmann/emotion-english-distilroberta-base
• Architecture: DistilRoBERTa transformer model
• Purpose: Detects emotional tone (anger, fear, joy, etc.)
• Features:
  • Chunked processing for long texts
  • Confidence-weighted aggregation
  • Sensationalism detection

4. Search Query Extraction (search_queries_service.py)

• Model: transformers pipeline for question generation
• Purpose: Extracts key claims from text for fact-checking
• Features: NLP-based query generation

Image Analysis Services

1. AI Image Detection (ai_image_service.py)

• Model: Custom EfficientNet-B3 [placeholder for url]
• Architecture:
  • EfficientNet-B3 backbone
  • Custom classification head
  • Binary classification (AI vs Real)
• Performance: 99% F1-Score
• Purpose: Identifies AI-generated vs real images
• Features:
  • 300x300 input resolution
  • Batch normalization
  • GPU acceleration

2. Face Detection (face_detection_service.py)

• Model: scrfd (pre-trained ONNX model)
• Purpose: Detects and locates faces in images
• Features:
  • Multi-scale face detection
  • Landmark extraction (eyes, nose, mouth)
  • Confidence scoring
  • NMS (Non-Maximum Suppression)

3. Deepfake Detection (deepfake_service.py)

• Model: Meso4 architecture (two variants)
  • Meso4_DF.h5: Trained on DeepFake dataset
  • Meso4_F2F.h5: Trained on Face2Face dataset
• Architecture:
  • 4 convolutional layers
  • Batch normalization
  • Max pooling
  • Dropout (0.5)
  • Sigmoid classification
• Purpose: Detects manipulated faces in images
• Features:
  • Dual-model ensemble
  • Face cropping integration
  • Confidence threshold adjustment

Fact-Checking Integration

Fact Check Service (fact_search_service.py)

• API: Google Fact Check Tools API v1alpha1
• Purpose: External verification of claims
• Features:
  • Multi-source fact-checking
  • Verdict aggregation
  • Confidence scoring
  • Source credibility assessment

📊 Model Performance

Model	Type	F1-Score	Dataset	Purpose
AI Text Detector	Custom Classifier	99%	Custom	AI vs Human Text
Fake News LSTM	LSTM Neural Network	99%	Custom	Fake vs Real News
AI Image Detector	EfficientNet-B3	99%	Custom	AI vs Real Images
Emotion Detector	DistilRoBERTa	N/A	Pre-trained	Emotion Analysis
Face Detector	SCRFD	N/A	Pre-trained	Face Detection
Deepfake Detector	Meso4	N/A	Custom	Deepfake Detection

🎬 Demo Media Files

The demo/ folder contains sample media:

Videos

Images

• demo1.png -
• demo2.png -
• demo3.png -
• demo4.png -
• demo5.png -
• demo6.png -
• demo7.png -

⚙️ Configuration

Environment Variables

# config.py
class Config:
    flask: FlaskConfig = FlaskConfig(
        SECRET_KEY = os.environ.get('SECRET_KEY', 'dev-secret-key'),
        UPLOAD_FOLDER = 'static/uploads',
        MAX_CONTENT_LENGTH = 50 * 1024 * 1024  # 50MB
    )

    service: ServiceConfig = ServiceConfig(
        FACT_API_KEY = "your-google-fact-check-api-key",
        FAKENESS_SCORE_THRESHOLD = 0.6,
        FACE_DETECTION_THRESHOLD = 0.5
    )

Model Paths

# Model file locations
AI_TEXT_DETECTOR = "./models/ai_text_detector.joblib"
FAKE_NEWS_DETECTOR = "models/fake_news_detector.pt"
EFFICIENTNET_AI_IMAGE = "./models/efficientnet_b3_full_ai_image_classifier.pt"
FACE_DETECTION = "models/face_det_10g.onnx"
MESO4_DF = "models/Meso4_DF.h5"
MESO4_F2F = "models/Meso4_F2F.h5"

🔗 API Endpoints

Method	Endpoint	Description
GET	/	Main analysis interface
POST	/analyze	Submit content for analysis
GET	/analysis/<id>	View analysis results
GET	/health	System health check

🤝 Contributing

1. Fork the repository
2. Create a feature branch (git checkout -b feature/amazing-feature)
3. Commit your changes (git commit -m 'Add amazing feature')
4. Push to the branch (git push origin feature/amazing-feature)
5. Open a Pull Request

Development Guidelines

• Follow PEP 8 style guidelines
• Add tests for new features
• Update documentation
• Ensure model compatibility

📝 License

This project is licensed under the MIT License - see the LICENSE file for details.

Acknowledgments

• Hugging Face for transformer models and pipelines
• Google for Fact Check Tools API
• PyTorch Team for deep learning framework
• OpenCV for computer vision utilities
• SCRFD for face detection models

Support

For support and questions:

• Create an issue in the repository
• Contact: [your-email@example.com]

Version History

v1.0.0

• Initial release with full multi-modal fake news detection
• 99% F1-scores on custom trained models
• Comprehensive web interface
• Integration with external fact-checking services

---

FactSight - Bringing truth to the digital age through advanced AI-powered analysis. 🛡️