JapanManufacturing2017 – ongoingTAIYO YUDEN
Deep-learning defect classification at Tamamura factory
Taiyo Yuden automated visual defect classification and extended it to overseas factories
Problem
Some defects were recognizable to human inspectors but difficult to classify accurately with conventional machine vision. Taiyo Yuden needed an alternative to rule-based image processing.
Approach
The Tamamura factory worked with Gunma Industrial Technology Center to evaluate deep-learning software and develop an inspection application. The partners compared detection performance and processing speed. Automated defect classification entered production in 2019 and subsequently spread to overseas factories.
Results
No published figures
Takeaway
The development team identified image selection and annotation as important work in approaching human classification accuracy.
Categories
- Size
- Large enterprise
- Industry
- Manufacturing
- Business function
- Production and quality
- Innovation type
- Process innovation
- AI technology
- Computer vision
- Adoption
- Partner co-development
- Evidence
- Company's own claim
Sources
Source summaries
Taiyo Yuden described joint research that automated defect classification previously performed visually. Production use began in 2019 and expanded overseas; subsequent research addressed software licensing costs through an open-source application.
Full details are in the original sources.