The challenge
A manufacturer of cosmetics and household products wanted to catch packaging defects (missing or misplaced labels, caps, wrong products) at line speed, and to make sure every carton and pallet carries the right code for the order being produced.
Our approach
- Scoped the pilot with the owner, the technical director, the AI & digital optimisation lead and IT.
- Designed the capture setup with two industrial cameras and a labelling tool so operators can build the training set quickly.
- Training a deep-learning model (PyTorch, CUDA) for bottle, cap and label inspection.
- Integrated production context from the EVOCON API and SAP exports, plus carton and pallet barcode readers.
Results & features
- Line and pallet screens show counters, defect rate, camera status and barcode checks in real time.
- Alerts when defects exceed an agreed threshold.
- Every decision is stored with its images for review; the operator keeps the final say.
