The big question:
How to minimize manual operations when the product is not homogenous?

Ukrainian startup produces eco product and wants to reach the capacity of 1,000,000 items daily. It is a challenge, because the quality checking and sorting processes are manual.

Being of natural origin, the items often have variable defects and require additional production cycle to make them ready for use.

Technology stack:

  • Engineering
  • Machine Learning
  • Computer Vision

Project Outcome

The Product:

Automated detector of defects

  • 5 types of defects inside the item and on its surface are recognized
  • Algorithm automatically sorts the items by the identified defect and collects analytics
  • 2 supervision points integrated into each assembly line
  • User interface shows items with the defects highlighted and general statistics

The Result:

99% of accuracy in product defect identification. Real-time automated quality check reduces the total production time at the factory.