Overview Deep Learning improves defect detection and reduces scrap rate on glass bottle inspection applications. Glass bottles are inspected at multiple angles at high speed for full coverage and to ensure the smallest of defects can be detected.
Not only is transparent glass bottle inspection a difficult imaging task, but also a difficult algorithmic challenge as there are a wide variety of defects types making it challenging for a traditional vision system and too fast for a human in-line quality inspector.
The Overview Solution
Our Deep Learning platform allows the vision system to meet high accuracy targets with minimal model training data required. Combined with multi-angle camera integration, each glass bottle is fully inspected for all defect types.
How does the
Deep Learning Work?
Easy integration combined with state of the art models
Flexibility in imaging bottles at multiple angles allows full coverage
Semantic segmenter identifies defect area for localization and sizing
Classifier categorizes defect type, allowing greater insight into defect rates by type
Models are easily trained for new defect types while maintaining performance
Full multi-angle inspection of glass bottles increases defect detection rates. Different defect types are separately flagged for better root cause analysis.
Demo the Snap Platform to see how we improve your factory efficiency