Founded in 1991 in Lviv, today ELEKS is a trusted Top 100 Global Outsourcing company that provides full-cycle software engineering outsourcing services, from ideation to finished products.
For over 30 years, the tech giant has been working as a software innovation partner with Fortune 500 companies, big enterprises, and technology challenges.
With a talent pool of 1600+ multiskilled specialists based in 12 offices distributed in key cities over three continents, ELEKS end-to-end software products help the world’s leading brands transform their businesses, increase their revenues, and operating costs.
Ukrainian BADER is the leading leather producer for the automotive industry. Established in Göppingen in 1872, the company develops and produces sophisticated natural material for more than 145 years as one of the leading international producers of premium leather for the automotive industry. With over 2000 specialists, BADER develops and manufactures Innovative interior solutions for mobile living environments, which are the first choice for automotive brands worldwide. The leading enterprises as Lamborghini, BMW, Audi, Lexus, Tesla, Toyota, and many others are the clients of BADER.
What was the challenge?
To enhance manufacturing productivity, BADER has a requirement to upgrade the quality control of leather production at a high level. So the ELEKS technical team started to automate the process of anomaly detection in the manufacturing process by implementing ML algorithms.
Moreover, at the very start of the project, there were no data to work with, so ELEKS team conducted also the process of Data Collection and Data Labelling for further models training.
What is the solution?
First of all, the engineering team researched the existing production process and implemented the technical solutions within it. The second step was to provide a flexible data model for verification, validation, and quality control of produced products, including search and reviewing results.
As a result – a model ensemble that can identify fastening details on the inner side of a seat and distinguish whether a defect is present. ELEKS team used Mask-RCNN + CNN stack for this purpose. To
- In the first stage, the Mask-RCNN model is trained to localize fastening details.
- In the second stage, the detected binary mask is passed to CNN trained on identifying a defect (binary: Defect or Non-Defect)
Special thanks to the ELEKS team for comprehensive support and assistance!
The project “Boosting local traditional industries with IT capacities” is implementing by TechUkraine in partnership with APPAU – Association of Industrial Automation of Ukraine, International Association Culture&Creativity Association with the financial support of the German government through GIZ that aims to increase the competitiveness of traditional industries in strong collaboration with Ukrainian IT companies.