🛍️ Purplle Store Intelligence

An edge-to-cloud computer vision and analytics platform built for the Purplle Tech Challenge.

🌐 Live Demo

Dashboard: https://store-intelligence-ai.onrender.com

API Documentation: https://store-intelligence-ai.onrender.com/docs

---

🌟 Key Features & Engineering Highlights

• Decoupled Edge-to-Cloud Architecture: Uses YOLOv8 and OpenCV at the edge to track visitors at 25+ FPS. Transmits telemetry as JSON to the cloud.
• High-Concurrency PostgreSQL Backend: Migrated from SQLite to containerized PostgreSQL utilizing SQLAlchemy 2.0 and SessionLocal dependency injection to eliminate 100% of database-locking bottlenecks during multi-camera telemetry bursts.
• Mathematical Correlation Engine: Computer Vision edge nodes frequently drop IDs. The API utilizes a custom "camera-aware" heuristic to rebuild disjointed customer journeys, recovering 100% of dropped tracking IDs to build perfect, cascading conversion funnels.
• Real-Time SPA Dashboard: An asynchronous Streamlit dashboard that isolates DOM updates via @st.fragment to auto-refresh metrics every 2 seconds without full-page UI flashes.

---

🏗️ System Architecture

CCTV → YOLOv8 Edge Tracker → Telemetry JSON → FastAPI → PostgreSQL → Mathematical Correlation Engine → Streamlit Dashboard

---

📂 Project Structure

📦 Store-Intelligence
├── app/
├── dashboard/
├── doc/
├── pipeline/
├── data/
├── tests/
├── Dockerfile
├── run.sh
├── requirements-api.txt
├── requirements-edge.txt
├── store_layout.json
└── README.md

⚙️ Technology Stack

Layer	Technology
Backend	FastAPI
Frontend	Streamlit
Database	PostgreSQL
Computer Vision	YOLOv8
Deployment	Docker
Hosting	Render
Language	Python 3.11

🚀 Local Setup

Clone Repository

git clone https://github.com/ayushsin9h/Store-Intelligence.git
cd Store-Intelligence

Create Virtual Environment

python -m venv venv

Windows:

venv\Scripts\activate

Linux/Mac:

source venv/bin/activate

Install Dependencies

pip install -r requirements-api.txt
pip install streamlit pandas requests

▶️ Run Backend

mkdir data
uvicorn app.main:app --host 0.0.0.0 --port 8000

API Docs:

http://localhost:8000/docs

▶️ Run Dashboard

streamlit run dashboard/web_dashboard.py

Dashboard:

http://localhost:8501

🎥 Running Edge Pipeline

pip install -r requirements-edge.txt

python pipeline/detect.py --source resources/Store_1/cam1.mp4

📊 Analytics

• Visitor Count
• Active Visitors
• Session Duration
• Conversion Funnel
• Queue Length
• Wait Time
• Zone Occupancy
• Dwell Time

🔄 API Endpoints

GET /health
POST /ingest
GET /analytics/funnel
GET /analytics/zones
GET /analytics/queue
GET /dashboard

🐳 Docker Deployment

docker build -t store-intelligence .
docker run -p 8000:8000 store-intelligence

☁️ Render Deployment

Uses:

• Dockerfile
• run.sh
• FastAPI
• Streamlit
• Auto Port Binding

🎯 Purplle Challenge Alignment

✅ Customer Journey Tracking

✅ Retail Conversion Analytics

✅ Queue Monitoring

✅ Zone Intelligence

✅ Multi-Store Visibility

✅ Edge-to-Cloud Architecture

✅ Real-Time Dashboarding

---

❤️ Built for the Purplle Tech Challenge

Transforming CCTV footage into actionable retail intelligence through Computer Vision, Analytics, and Cloud Engineering.