Real-Time Quality Control with Computer Vision and Edge AI

A leading Fast-Moving Consumer Goods manufacturer partnered with Darwin Edge and the Bonseyes consortium to streamline quality control on the production line. Through a Deep Learning-based Computer Vision model optimized for edge deployment, we enabled real-time defect detection and categorization, improving production speed, reducing waste, raising product quality and ultimately improving profitability

Industrial conveyor system with high-precision cameras for real-time quality control, leveraging computer vision to detect defects in manufactured products.

Table of content

Introduction

A large FMCG (Fast-Moving Consumer Goods) manufacturer aimed to enhance production line quality control by implementing automated quality checks directly on the line. Real-time detection was crucial for effective quality control, as delays in finding defects would increase waste, reduce production efficiency and impact profits.

Solution for Real-Time Quality Control

Darwin Edge and the Bonseyes consortium created a Deep Learning-based Computer Vision model specifically designed for detecting and categorizing defects in real time. This solution helped the manufacturer improve production speed, reduce waste, raise product quality and ultimately improve profitability. Optimized for Edge AI, the model operates directly on production line machines without relying on cloud connectivity, ensuring no latency. This setup allows immediate defect detection and real-time alerts, enabling operators to respond quickly.

Three bottles showcasing computer vision-based quality control. One bottle is correctly labeled, while two show label defects, highlighting the use of computer vision for defect detection.
Example of automated quality control using computer vision to detect product label defects in real-time.

Technology Stack

We used a DigiFlex camera setup to capture images of defects, which were then labeled manually for training. Using advanced Deep Learning algorithms for image classification and segmentation, the models were built on a Python-based Keras/TensorFlow backend and trained on AWS Sagemaker, ensuring scalability.

Diagram of a neural network architecture used in computer vision, showing input, hidden, and output layers, applied to quality control in real-time manufacturing.
Diagram of a neural network used in computer vision to enable accurate defect detection on the production line.

Results and Benefits of Automated Quality Control

Darwin Edge, in collaboration within the Bonseyes consortium, delivered a working prototype that confirmed automated quality checks could be achieved at scale and in real time. Key achievements included:

  • Defect Detection and Categorization: Our Deep Learning-based Computer Vision model was trained on over 5,000 images to accurately identify and categorize defects, supporting high-precision quality control.
  • Scalability: The prototype was deployed in an R&D lab, showing that the model can handle large-scale, real-time quality checks.
  • Improved Quality Assurance: Automated defect detection reduced the time and effort of manual inspections, especially for complex defects, lowering the chances of human error.

Partner with Darwin Edge to elevate your production line with cutting-edge quality control solutions that improve efficiency, reduce waste, and enhance product quality in real time.

Subscribe to Our Monthly Newsletter

Subscribe to our newsletter to receive the latest news, insights, case studies, and innovations from Darwin Edge directly in your inbox. Stay informed with updates on our cutting-edge Edge AI solutions, upcoming events, and industry trends.