× L-Labs «Pizzafabrika × True Detection: quality standards protected by AI
How we trained a neural network to evaluate every pizza across 54 restaurants automatically and in real time
A problem that cannot be solved manually
When a chain grows to dozens of locations, maintaining a single quality standard becomes increasingly difficult. Checks are selective, responses are delayed, and every inspector may evaluate quality differently.
It is physically impossible to check every pizza manually, which means violations can reach the customer unnoticed.
The team often learns about a violation only after a customer complaint — hours later, when it is already too late to fix anything.
Each inspector evaluates products in their own way, causing standards to become inconsistent from one location to another.
How the system works
The system was integrated directly into the preparation process. No extra actions are required from kitchen staff.
A camera in the kitchen automatically takes a photo of every finished pizza. The cook continues working as usual.
The photo, together with order data, is sent to the server and processing system in a fraction of a second.
The neural network checks the baking color and evenness, shape, slicing, toppings, recipe compliance, and other parameters.
If a violation is detected, the system instantly sends a message to the Telegram channel of the specific restaurant.
All data is available in the web dashboard. At the end of each day, week, or month, automatic Excel reports are generated with a breakdown by location.
15 quality parameters
The system is trained to recognize all typical violations of PizzaFabrika standards.
Three stages, no unfinished launch
A gradual approach allowed us to deliver a fully operational tool to the client, rather than a beta version.
We launched the system in pilot mode. The model was trained on archived photos and tested in real kitchen conditions.
We refined the model for PizzaFabrika’s recipes, lighting conditions, and presentation standards. False positives were reduced, and accuracy was improved.
The system was rolled out across the entire chain as a proven, stable tool — with no surprises after launch.
What changed for the business
The main value is not automation for its own sake, but the tangible improvements it brings to the chain’s daily operations.
Previously, only a sample of orders was inspected. Now every single pizza is checked without exception.
A violation is detected at the preparation stage, and the notification is sent immediately.
Standardized apply consistently across all 54 restaurants.