How Krevera Improved Defect Detection with Labellerr AI

As automated inspection systems continue to evolve across manufacturing and logistics industries, detecting small physical defects on industrial equipment and storage materials has become increasingly important.

Krevera partnered with Labellerr to build a high-quality image annotation pipeline focused on identifying defects in plastic storage bins, commonly known as totes. These defects ranged from subtle scratches and worn-out surfaces to dents and structural damage that were often difficult to detect visually.

To support the development of reliable computer vision models, Krevera required precise annotations across thousands of images every month while maintaining strict quality standards and operational consistency.

About the Client

Krevera works on industrial inspection and defect detection workflows involving plastic storage bins used across warehouse and logistics operations. Their objective was to improve automated defect identification systems capable of recognizing even the smallest surface-level damages on totes.

Because many of these defects were extremely subtle and inconsistent in appearance, the success of their AI models depended heavily on highly accurate and detailed image annotations.

The Challenge

Detecting defects on plastic totes presented several unique annotation challenges:

• Defects were often extremely small and visually difficult to identify
• Surface damage varied significantly in shape, texture, and severity
• Scratches and dents could appear under inconsistent lighting conditions
• Some defects blended naturally with the tote surface material
• Maintaining annotation consistency across large monthly datasets was critical

Even minor inaccuracies in annotation could reduce model performance during training, especially for small-object defect detection tasks.

The Solution

Labellerr provided a dedicated image annotation workflow tailored specifically for industrial defect detection use cases.

The project focused on annotating:

• Surface scratches
• Worn-out regions
• Structural dents
• Other visible tote defects

Using Labellerr’s labeling platform, the annotation team ensured:

• Precise defect localization
• Consistent annotation quality
• Efficient monthly delivery cycles
• Smooth collaboration and review workflows

The platform enabled streamlined project management while supporting accurate annotation of very small and difficult-to-detect defects.

Scale of Operations

The engagement operated on a continuous monthly annotation cycle.

Key project highlights included:

• Annotation of approximately 5,000 images every month
• Consistent delivery based on client dataset capacity
• Stable long-term annotation operations
• Reliable quality assurance processes throughout the workflow

The combination of trained annotation teams and Labellerr’s annotation infrastructure ensured that every monthly batch was delivered efficiently and with high accuracy.

Results

Through the collaboration, Krevera was able to maintain a reliable dataset generation pipeline for their defect detection initiatives.

The partnership helped achieve:

• High-quality annotations for small defect detection
• Consistent monthly dataset processing
• Smooth annotation operations at scale
• Improved dataset reliability for model training
• Efficient handling of challenging industrial inspection images

By combining operational consistency with precise annotation standards, Labellerr helped support Krevera’s computer vision workflows for automated tote inspection and defect analysis.

Conclusion

Industrial defect detection systems rely heavily on accurate and detailed training data, especially when defects are subtle and difficult to identify visually.

By supporting large-scale monthly annotation operations for tote defect detection, Labellerr enabled Krevera to build and maintain reliable datasets for computer vision model development while ensuring annotation quality and operational efficiency throughout the project lifecycle.

‍