AI-Powered Defect-Free Packaging System

Use Case:

Automated object recognition to eliminate packaging defects in high-value FPV goggle shipments improving customer satisfaction and reducing costly reshipments and returns.

Outcome:

• 98.02% mean average precision (mAP) in object classification;
• Zero packaging defects across all test runs;
• Real-time user interface for packaging line workers;
• Significant cost savings on missing item reshipment (estimated €3,000 per 1,000 units).

Ecosystem Support:

StairwAI support including technical mentoring, AI expert services, and computing resources for dataset creation and model training.

AI Relevance:

The ORQAI pilot highlights the impact of AI democratization for SMEs through:

1. low-barrier AI integration in traditional manufacturing processes;
2. Use of domain-specific, small-scale datasets;
3. AI-empowered quality assurance without changing legacy infrastructure;
4. Replicable blueprint for defect prevention in packaging lines.

Summary:

Orqa d.o.o., a Croatian SME renowned for its FPV goggle systems, faced a recurring issue in its packaging line: incorrectly assembled packages due to missing or miscounted components. To address this, Orqa developed ORQAI, an AI-powered packaging assistant, within the StairwAI project framework. The solution integrated computer vision and a web-based AI inference system to automatically verify the presence and correctness of each item placed in a package.
The pilot involved building a curated dataset of over 1,000 annotated images, training several models, and selecting the most accurate configuration. The final model, achieving 98.02% mAP, was deployed on a real production line, allowing a mounted camera to identify all package components in real-time. A user interface displayed recognized items to operators, providing immediate feedback and preventing errors before sealing the packages.
Beyond technical excellence, ORQAI demonstrated tangible business impact. By avoiding packaging defects, Orqa reduced reshipping and replacement costs, while improving customer trust and satisfaction. The pilot also established a framework for future product integration: new models can be trained in-house using the AI workflow and extended to other packaging stations.
StairwAI support proved crucial: access to an image processing expert and mentoring enabled a smooth AI transition. With TRL 7/8 achieved, ORQAI stands as a scalable, AI-powered quality control solution ready for commercial deployment and adaptation across similar high-precision packaging environments.