We have confirmed the effectiveness of detecting defective castings and would like to introduce cases where detection accuracy has significantly improved!
To prevent defective products from flowing into the next process, it is necessary to conduct a complete visual inspection before shipment. However, there were issues such as individual differences among inspectors and missed detections due to fatigue.
In response, we turned to an "algorithm." This is a system that allows humans to instantly understand what objects are in an image, where they are located, and how they interact with each other.
Given that the ability to detect objects at a glance is a strength, we believed we could conduct inspections and decided to challenge this method.
As a result, we confirmed the effectiveness of detecting defective areas in castings, and by evaluating a suitable AI model, we were able to use it with confidence on-site. By leveraging our company's know-how to suppress disruptive factors, we significantly improved detection accuracy.
Additionally, even after applying it on-site, we could add unknown images to the training data, allowing us to provide feedback for improving detection accuracy.
[Background]
- To prevent defective products from flowing into the next process, it is necessary to conduct a complete visual inspection before shipment.
- There are issues such as individual differences among inspectors and missed detections due to fatigue.
- If any problems arise afterward, there is no physical product available for additional verification.
*For more details, please refer to the PDF document or feel free to contact us.*