Detecting Filling Defects in Cosmetic Products with Computer Vision
In the cosmetics industry, quality is not a detail — it is the product.
A poorly filled bottle, a cream below the declared level, or a visible air bubble can undermine customer trust and damage a brand’s reputation.
Today, AI-powered Computer Vision is transforming quality control, making it more accurate, scalable, and automated than ever before.
Why Is Fill-Level Control So Critical?
In the beauty world, packaging is part of the experience. An inconsistent fill level can lead to:
Customer complaints and returns
Regulatory issues
Brand reputation damage
Waste and rework costs
Manual inspections, besides being slow and prone to human error, are not sustainable on high-speed production lines.
This is where Computer Vision with AI comes in.
What Is Computer Vision in Quality Control?
Computer Vision is a branch of artificial intelligence that enables systems to “see” and interpret images or video.
In industrial environments, high-resolution cameras installed along the production line capture images of products, while Deep Learning algorithms analyze each unit in real time.
The goal? Automatically identify anomalies such as:
Fill level too low or too high
Presence of air bubbles
Non-uniform filling
Leakage or overflow around the bottle neck
How Does Filling Defect Detection Work?
Image Acquisition
Industrial cameras capture images of the product immediately after filling.
Pre-processing
The system normalizes lighting, contrast, and orientation to ensure consistency.
AI Analysis
Deep Learning models — often based on Convolutional Neural Networks (CNNs) — analyze:
The upper liquid boundary
Comparison with an ideal fill threshold
Abnormal patterns within the content
Automated Decision
If a defect is detected, the system can:
Trigger an alert
Activate an automatic rejection mechanism
Log data for statistical analysis
All of this happens in milliseconds.
Key Technologies Used
2D and 3D vision systems for more accurate volumetric measurements
Controlled lighting (backlight or structured light) to enhance liquid-level detection
AI frameworks such as TensorFlow and PyTorch for model training
Integration with industrial PLC and MES systems
Benefits for Cosmetic Manufacturers
Higher Accuracy
Significant reduction of false positives and false negatives compared to manual inspection.
Speed
100% product inspection, even on high-throughput production lines.
Traceability
Data collection for predictive analysis and continuous improvement.
Cost Reduction
Fewer rejects, fewer returns, less manual intervention.
Challenges to Consider
It’s not entirely plug-and-play. Common challenges include:
Variations in packaging transparency
Reflections on glossy bottles
Differences in product viscosity
The need for well-balanced datasets during model training
An effective AI system requires a solid data collection phase and continuous performance monitoring.
Practical Use Case
Imagine a filling line for transparent facial serum bottles:
The camera detects the liquid level
The AI model identifies air bubbles exceeding a defined threshold
The system automatically rejects defective units
Data is logged to identify recurring issues in the filling machine
Result: consistent quality and full process control.
The Future: Toward Predictive Quality Control
The next step is not just detecting defects — but predicting them.
By integrating Computer Vision, AI, and production data analytics, manufacturers can:
Identify patterns that anticipate filling errors
Optimize process parameters in real time
Reduce downtime and unplanned maintenance
This shifts quality control from reactive to intelligent and adaptive.
Conclusion
The adoption of AI-powered Computer Vision for detecting filling defects in cosmetic products is no longer futuristic — it is a concrete competitive advantage.
In a market where aesthetics and perceived quality are essential, automating inspection means protecting the brand, improving efficiency, and ensuring that customers receive a flawless product.
Want to know more? Contact us at info@metalya.it