- AI
- NVIDIA
- Case story
Discover how to prevent mislabeling in manufacturing and boost productivity. In this case study we share the benefits of implementing machine vision technology and IoT (Edge AI) for optimizing label inspection in a luxury candy factory.
Our label inspection solution is ideal for factories across various industries, where products require accurate labeling. With a combination of cutting-edge technology and existing technical infrastructure, we deliver efficient and precise label inspections. Save time and ensure quality compliance, guaranteeing smooth production processes and customer satisfaction.
Business Goal
A large Danish candy factory aimed to streamline its production process by leveraging artificial intelligence for label inspection. The existing machine vision system was limited to processing only 12 candy jars per minute. To overcome this bottleneck, we sought to enhance the system's capabilities.
Our initial objective was to implement an emblem notification system that would signal label machine approval and introduce even faster cameras to double the processing rate. The focus was to accurately detect and verify if the label was applied in alignment with the candy jar emblem. These improvements could be integrated seamlessly without interrupting ongoing operations.
Subsequently, we planned to train the system to read labels and comprehend their content, including language, product details, and placement accuracy. As a final step, we envisioned automating the removal of defective products from the production line, with the system alerting operators to any irregularities.
Solution
To address the client's need for enhanced label inspection, we developed a machine vision solution, powered by YOLO (You Only Look Once) which offers a game-changing approach to label inspection in food manufacturing.
Key components of the solution:
-
High-speed Basler camera:
The integration of a Basler camera enabled a significant increase in processing speed, allowing us to inspect up to 30 candy jars per minute, more than double the previous capacity of 12 jars per minute. -
Containerized architecture:
The system was deployed in a containerized environment, allowing for efficient scaling and management. Three primary containers were developed:
Inference container: Powered by Python and a supervision framework, this container houses the AI model responsible for real-time label inspection.
Model container: This container stores the trained deep learning model, which is optimized for performance and accuracy.
Web interface container: A simple web interface was developed to provide operators with real-time adjustment capabilities and insights.
Result
Our machine vision-based label inspection solution had a transformative impact on the candy manufacturing plant's operations. By automating the inspection process, we achieved:
- A remarkable 150% increase in productivity, enabling the inspection of 30 candy jars per minute compared to the previous rate of 12 jars per minute.
- Faster identification of properly labeled candy jars, reducing inspection time by 60%.
- Higher percentage of high-quality products reaching the market.
- Improved overall manufacturing workflow, optimizing resource allocation and contributing to a more efficient and cost-effective production line.
Challenges
Optimizing for high-speed inspection without compromising accuracy
One of the main challenges we encountered was achieving a balance between speed and accuracy during the label inspection process. Our team fine-tuned the computer vision algorithms and hardware setup to ensure fast and reliable identification and validation of labels without compromising quality.
Adapting to multilingual and international requirements
Considering the diverse customer base of the luxury candy manufacturing plant, it was important to implement a system that could efficiently sort labels based on languages and countries. Our solution integrated language detection algorithms and country-specific label categorization to effectively meet these requirements.
Streamlining production and removing defective products
In addition to label inspection, our solution included an automated mechanism to remove defective products from the production line. This further enhanced the overall production efficiency by ensuring that only high-quality candy jars reached the packaging and distribution stages.
According to Worldmetrics AI can help improve food safety by detecting contaminants is real time with 99% accuracy.
