Precision Customer Analytics through Computer Vision
Retail Flow is an intelligent computer vision system designed to transform physical movement into actionable retail intelligence. By utilizing YOLO machine learning models, the system tracks customer behavior, measures zone density, and correlates foot traffic with sales performance.
University: Universidad Europea
Professor: Alejandro Perdiguero
| Name | Role | Position |
|---|---|---|
| María Parra Ramos | Software Engineer | CEO |
| Santiago Campos Snels | Data Engineer | Project Manager |
| Carlota Prósper Hernández | Data Engineer | Worker |
| Javier Hernández Pérez | Data Scientist | Worker |
| Vicente Vazquez | Data Scientist | Worker |
| Ashley Harris Velasquez | Data Analyst | Worker |
- 🔍 Object Detection: Real-time boundary box prediction to isolate human clients from store fixtures.
- 📈 Traffic Analysis: Granular tracking of user density and dwell time per zone.
- 💡 Business Intelligence: Bridging the gap between "interest" (browsing) and "conversion" (purchasing).
The system monitors a 100 ft x 50 ft floor plan, strategically divided into five functional zones:
- Entrance: Control point for total conversion rate calculation.
- Sales Floor: Primary data source featuring high-engagement shelving units.
- Fitting Rooms: Lounge and trial area to analyze product attraction.
- Self-Checkout: Final transaction point for sales correlation.
- Storage Room: Restricted administrative zone.
We deploy a specialized 5-camera matrix to ensure zero blind spots and high-fidelity tracking:
| Hardware | Location | Field of View | Primary Goal |
|---|---|---|---|
| Bullet Camera | Entrance | 90° (30ft) | Entry/Exit Counting |
| Fisheye Cam A | Sales Floor | 360° (35ft) | Path Tracking & Heatmaps |
| Fisheye Cam B | Sales Floor | 360° (35ft) | Path Tracking & Heatmaps |
| Dome Camera | Fitting Rooms | 90° (30ft) | Dwell Time Measurement |
| Turret Camera | Checkout | 70° (15ft) | Transaction Correlation |
The project utilizes a robust relational SQL database structured to handle high-frequency detection data:
graph TD
Store --> Zones
Store --> Sales
Zones --> Cameras
Zones --> Traffic
Cameras --> Detections