The challenge

During the vehicle manufacturing process, SEAT develops the designs for each of the parts that make up the vehicle. These designs are sent to the manufacturers for production. A critical element in this process is the Drückbild, i.e. the contact area of a part with the other parts with which it must fit or adhere.

In some cases, discrepancies may arise between SEAT and the manufacturers about the validity of the Drückbild. Managing these differences in an agile and efficient manner is key to guaranteeing the quality of the final product and meeting production deadlines.

Objectives

The aim is to establish an objective method to accurately quantify the percentage of the manufactured Drückbild that complies with the designs provided by SEAT.

How have we done it?

• Object Segmentation
• Hugging Face
• Segformer
• Computer Vision
• Active Learning
• AWS

Results provided

• Transparent and consensual validation between manufacturers and SEAT.
• Anticipation and correction of deviations in early stages of the production process, thus ensuring better overall quality and efficiency.