SPORTS GUARDIAN is a high-performance, anti-piracy ecosystem designed for sports media rights holders. It provides a "Mission Control" interface to register protected assets, deploy scraping nodes across social platforms, and identify copyright infringements using multi-modal fingerprinting (Visual, Audio, and Scoreboard OCR).
- Ingestion: Upload master sports media (clips, match highlights).
- Multi-Modal Hashing: Generates
pHash(Perceptual) andPDQ(Geometric) vectors for visual matching resistant to resizing and cropping. - Acoustic Fingerprinting: Audio-based detection for streams with obscured video.
- Distributed Scraping: Nodes targeting YouTube, Instagram, and Reddit using
yt-dlpand official APIs. - Real-time Discovery: Search-based discovery (e.g., "Live Match Stream") with frame-buffer analysis.
- Live Terminal: Real-time feedback of scraping and analysis progress.
- Scoreboard OCR: Uses YOLO-based detection (RT-DETRv2) to sync match time/score with the master asset.
- AI Moderator: Agentic reasoning using Gemini Pro (
google/gemini-pro-latest) to analyze metadata and piracy signals. - Human-in-the-Loop: A dedicated review queue for moderate-confidence detections.
- Notice Generation: Automated creation of copyright infringement notices.
- Blockchain Proof: Immutable anchoring of asset fingerprints for legal enforcement.
- OAuth2 with Password Flow: Secure login using
FastAPIandJOSEJWT tokens. - Role-Based Access: Protected API endpoints and frontend guards to ensure only authorized operators can manage assets and scans.
- Bcrypt Hashing: Industry-standard password encryption for user security.
- Token-Based Interceptor: Angular interceptor automatically attaches JWT tokens to outgoing API requests for seamless, secure communication.
Overview of protected assets, active scan jobs, and threat levels.
Securely upload and fingerprint master media assets.
Initiate and monitor real-time scraping operations across social platforms.
Detailed breakdown of matches found, including similarity scores and visual comparisons.
Operators can manually verify flagged content before enforcement actions.
Automated generation of copyright notices and takedown reports.
| Component | Technologies |
|---|---|
| Backend | FastAPI, SQLAlchemy, LangGraph, OpenCV, ImageHash, PDQHash, yt-dlp |
| Frontend | Angular 19+, Signals, Standalone Components, Neo-Brutalism UI |
| CV Model | RT-DETRv2, Ultralytics, FastAPI (Microservice) |
| Database | PostgreSQL, Alembic |
| AI/LLM | Gemini Pro (google/gemini-pro-latest), Ollama |
- Python 3.12+
- Node.js 20+
- PostgreSQL
- FFMPEG (for media processing)
cd backend
# Install dependencies using UV
uv sync
# Configure Environment
# Create a .env file based on the requirements
# DATABASE_URL=postgresql://user:password@localhost/shield_media
# Run Migrations
alembic upgrade head
# Start API Server
fastapi dev app/main.pycd frontend
npm install
npm startAccess the UI at http://localhost:4200
cd model
pip install -r requirements.txt
python main.pyRuns on http://localhost:8001 by default.
- Login: Access the Mission Control terminal via the Login Page.
- Register Asset: Go to "Register Asset" and upload your master match clip. The system will automatically generate visual and audio fingerprints.
- Start a Scan: Navigate to "New Scan", enter search queries (e.g., "Team A vs Team B Live"), and select the platforms to monitor.
- Monitor Progress: Use the Scan History to track active jobs in real-time.
- Review Detections: Check the "Human Review" queue for any content flagged with a
REVIEWstatus. - Take Action: For
FLAGGEDcontent, view the Scan Result and generate an automated Copyright Notice.
├── backend/ # FastAPI Application & Business Logic
├── frontend/ # Angular 19+ Neo-Brutalism UI
├── model/ # RT-DETRv2 Object Detection Service
├── screenshot/ # Application Screenshots
└── Detailed_Technical_Documentation.md # Full architecture deep-dive
Internal Project - All Rights Reserved.