Skip to content

Latest commit

 

History

84 Commits

Folders and files

NameName
Last commit message
Last commit date
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 

Repository files navigation

⚽ VisionPlay AI: Tactical Graph Intelligence Engine

VisionPlay AI is an elite-tier tactical intelligence platform designed for high-stakes football analytics. Unlike traditional statistics that focus on individual actions, VisionPlay AI models the pitch as a dynamic spatio-temporal graph, quantifying the invisible architecture of team coordination and defensive stability in real-time.


🚀 The Winning Strategy

VisionPlay AI transforms raw video footage into actionable structural insights. By analyzing the "structural health" of a formation using advanced graph theory, the platform allows analysts to identify tactical fractures and lynchpin players before they impact the scoreboard.

🧠 Core Algorithms & Tactical Meaning

Algorithm Strategic Intelligence
Graph Laplacian Formation Entropy: Measures the "chaos" vs. "order" of the team structure.
Tarjan’s Algorithm Lynchpin Isolation: Identifies the single "Articulation Point" player whose neutralization fractures the team.
Floyd-Warshall Team Diameter: Audits global compactness and spacing in real-time.
Spectral Entropy Stability Index: Predicts the probability of defensive collapse during transitions.

🛠️ Tech Stack

  • Frontend: Next.js 15 (App Router), Bun Runtime, Tailwind CSS, Framer Motion (Cinematic UI), Recharts (Data Visualization).
  • Computer Vision: Python 3.10+, YOLOv8 (Object Detection), OpenCV (Frame Processing).
  • Mathematical Engine: NetworkX (Graph Theory), NumPy, SciPy.
  • Backend API: FastAPI (High-performance WebSocket streaming).

🕹️ Platform Modules

1. Command Center (Dashboard) 🖥️

The nerve center of the platform. Features a Live Player Intelligence Engine table that cross-references YOLO kinematics with graph metrics.

  • Real-time Metrics: Speed, Fatigue, Synergy Index, and Decision Quality.
  • Predictive Analytics: Automated risk assessment for structural fractures.

2. Live Engine 📡

A high-fidelity tactical HUD that processes match footage in real-time.

  • ** HUD Overlays**: Team classification (Green vs. White) and unique Node IDs.
  • Temporal Stability Analysis: A dynamic timeline tracking Formation Entropy with variance stretching for precise anomaly detection.

3. Forensic Replay Lab 🔍

A tool for post-match audits that allows analysts to scrub through footage with frame-accurate precision.

  • Automated Event Detection: Scans the timeline for entropy spikes and articulation vulnerabilities.
  • Audit Reports: Instant generation of tactical intelligence reports based on AI-identified critical moments.

4. Tactical Sandbox (Simulations) 🏗️

An interactive environment where analysts can drag-and-drop players to test "what-if" scenarios.

  • Live Graph Feedback: Formation Entropy updates in real-time as nodes are moved.
  • Ghost Formations: Compare current positioning against the "original" YOLO-detected tactical setup.

5. Intelligence Reports 📄

A deep-dive audit tool for structural symmetry and formation flow.

  • Laplacian Eigenvalue Timelines: Visualizes the stability of the match over time.
  • Articulation Node Audit: A historical breakdown of which players acted as structural lynchpins.

⚡ Developer Setup

1. Frontend (Next.js + Bun)

# Install dependencies
bun install

# Start the dev server (includes automatic port cleanup)
bun dev

2. Vision Pipeline (Python)

cd visionplay-pipeline
# Create virtual environment
python -m venv .venv
source .venv/bin/activate

# Install requirements
pip install -r requirements.txt

# Run the detection pipeline to generate tactical data
python src/main.py

📁 Project Structure

├── app/                  # Next.js App Router (Dashboard, Live, Replay, etc.)
├── components/           # Reusable UI components (Navbar, Glassmorphism cards)
├── public/
│   ├── data/             # AI-generated tactical_data.json
│   └── videos/           # Processed YOLO output videos
└── visionplay-pipeline/  # Core AI Engine
    ├── src/              # YOLO & Graph algorithms
    ├── input_videos/     # Raw match footage
    └── models/           # YOLOv8n weights

Built with passion for the GDG Hackathon by Noah Menezes.

About

Resources

Stars

1 star

Watchers

0 watching

Forks

Releases

Packages

Contributors

Languages