Manually inspecting millions of electric seat motors requires thousands of workers globally. Defects on the Commutator bar which makes up part of the motor are very common.
Automated defect detection reduced the man power required and therefore reduced the associated costs substantially. It also allowed staff to focus on other business critical tasks.
- Set up cameras to take images of the commutator bar surfaces
- Identify and label the different types of defects
- Develop a system to classify images as containing defects or not.
- Using Generative Adversarial Networks (GANs) to develop anomaly detection and defect classification models
- Combine these models to create a defect detection application