Segmentation and classification, pharma
OEM (original equipment manufacturer)
Improve the current quality control system on contact lens production lines, reducing the false scrap rate and enabling the accurate localisation of defects to facilitate further customer quality analysis.
Some of the customer’s inspection issues are due to the lack of accuracy of the current vision system on the line – which checks all the contact lenses – with a single lens inspection cycle time of 144 hundredths of a second. The current system, in particular, generates a high percentage of false rejects, due to the extremely strict acceptance thresholds set by the operators as a safety measure. Variable production conditions and the need for regular and complex fine-tuning make the present solution not completely efficient.
A model for the automatic detection of defects in contact lenses was trained using AI-go Studio, providing an output segmentation of the anomalies so that it is possible to trace the areas of attention on which the algorithm focused and, if necessary, to support further quality analysis.
Thanks to AI-go Runtime, the model could be deployed in production, after the existing computer vision system, so that the two systems could work in tandem and guarantee double quality control.
Increased OEE (Overall Equipment Effectiveness). Maintaining the same quality assurance, but reducing false rejects.
Better anomaly detectability– even in sub-optimal visual conditions.
Data-driven approach. To keep track of defects in order to review the process and/or supply chain for continuous improvement.
Quality evaluation objectification. The quality evaluation is more reliable over time.