Computer-Vision Inventory Tracking System for a High-Volume Bakery

Overview

Built a computer-vision inventory system for a high-volume bakery that tracks shelf and case stock levels throughout the day using cameras already installed for security, avoiding the need for staff to manually count product during busy shifts. The system flags when a popular item is running low so staff can start a new batch before it actually sells out, and it logs end-of-day waste by comparing what was baked against what was sold or discarded. This gave the bakery real production data it never had before, replacing guesswork-based baking schedules built on the owner's intuition. The system was tuned specifically to recognize the bakery's own product line rather than using a generic retail object-detection model. A high-volume bakery relied on the owner's intuition for baking schedules, with no real production data, leading to both stockouts of popular items and significant unsold waste on slower days. We built a computer-vision system using the bakery's existing security cameras to track shelf and case stock levels throughout the day, flagging low stock before items sell out and logging end-of-day waste automatically. We trained the detection model specifically on the bakery's own product photos rather than a generic retail dataset, then validated stock-level accuracy against manual counts over a two-week pilot before relying on it for baking decisions. End-of-day unsold product waste dropped by an estimated 20%, with staff now working from real-time stock visibility instead of periodic manual counts and guesswork-based baking schedules.