BelgiumRetail and e-commerce2019 – ongoingColruyt Group
Deep-learning recognition of loose fruit and vegetables at the till, built with Robovision
Colruyt's camera model picked the right fruit or vegetable on the tablet in 97% of early test cases
Problem
Cashiers at Colruyt had to enter produce codes by hand for loose fruit and vegetables. That slowed the till.
Approach
In August 2019 Colruyt began a three-month test in its Kortrijk store, the first Belgian supermarket to try this. A camera above the scales photographs the produce and an AI model developed with Robovision names it. The customer confirms the choice on a tablet. The model covers the 120 fruit and vegetable types Colruyt stocks. The cashier can scan other items while the customer weighs produce.
Results
Takeaway
Colruyt kept a human confirmation step on the tablet even with a high first-test hit rate. A later Colruyt scale test, reported in 2025, suggests the idea was still being tried.
Categories
- Size
- Large enterprise
- Industry
- Retail and e-commerce
- Business function
- Production and quality
- Innovation type
- Process innovation
- AI technology
- Computer vision
- Adoption
- Partner co-development
- Evidence
- Press reporting
Sources
Source summaries
The article describes a three-month test of AI produce recognition at Colruyt in Kortrijk, built with Robovision. It reports that the right product showed on the tablet in 97% of early tests and that the model knows 120 types. It does not report what happened after the test.
The article says Colruyt was testing a smart checkout scale that identifies loose fruit and vegetables, alongside its camera rollout. It gives no result for that test.
Full details are in the original sources.
