Final Year Project

Project Title

FMCG Shelf Monitoring

Description

Retail shelf monitoring remains largely manual — brand representatives physically inspect stores while surveillance teams watch camera feeds. Our system automates this using YOLOv8 with a novel business-rules-first approach grounded in consumer psychology research. Unlike existing solutions that simply count products, we define "detected" through three criteria: 70% face visibility, two-row reachability, and correct orientation — measuring purchasability, not just presence. Validated through field research at Springs Supermarket and approved by Carrefour and Al-Fatah via our industry partner Multinet, the system achieves 97.2% mAP and generates automated severity-based restocking alerts. Built as a full-stack application with Next.js, FastAPI, and Supabase, it reduces response time from hours to minutes, delivering real-time shelf compliance data previously unavailable to FMCG brands.