Computer Vision reduces quality inspection lead time by 80%

Aiming for industry-best standards

The German manufacturer is a global company with factories across Europe, Mexico, China and India. It was striving to manage a high standard of quality for its factory equipment. Although it had a global team, manual quality inspection was expensive and time-consuming with a direct impact on the business bottom line.

Envisioning a smarter solution

Eviden designed Computer Vision-based models for an automated quality inspection of the products at the client’s factory locations.

Tensorflow models were deployed with the inference of camera feed on Edge for better performance. Leveraging these CCTV cameras installed at various angles, the solution can inspect products delivered from the assembly lines, including communication devices, chipsets, bolts and nuts, motors, printed circuit boards and generators. An OCR scan of equipment text images enables employees to add meta data with autoclassification to identify any defects.

This containerized, microservices architecture also has an in-memory database and lean UI console to deploy E2E framework on each Edge at the factory sites. Additionally, employees can be trained with modules on the cloud to enable parallel and distributed training on GPUs while deployment on Edge locations.

Automated inspection, accelerated efficiency

Eviden’s solution has delivered an 80% reduction in quality inspection lead time with more than 85% accuracy of model prediction. Overall, this has increased the manufacturer’s process efficiency.