Plunger 결함 검출을 위한 생성형 AI 기반의 데이터 증강
Data augmentation with Generative AI for Plunger Defect Detection
1
조회
0
다운로드
0
인용
0
참고문헌
키워드
초록
최근 제조 산업 분야에서 AI를 도입해 제품의 결함을 분석하고 있다. 이때, AI를 통한 비전 검사에서 무엇보다 중요한 것은 양질의 데이터를 충분히 확보하는 것이다. 그러나 제조 데이터 취득을 위한 시스템을 구축하기 쉽지 않고 불특정한 결함을 가지는 데이터는 수집하는 데 더 많은 시간이 소요된다. 본 논문에서는 생성형 AI 기반의 제조 데이터를 통한 데이터 증강 비전검사 기법을 제안한다. 고품질 데이터 획득을 위해 Stable Diffusion의 입력 데이터와 최적의 파라미터 조합을 찾아 데이터를 증강하였으며, 이를 바탕으로 비전검사를 실행한 결과 우수한 정확도를 보였다. 이러한 접근 방법은 데이터 수집의 어려움을 극복하고, 제품 품질 향상에 기여할 것으로 기대된다.
Abstract
The manufacturing industry has recently adopted AI to analyze defects in products. The most important aspect of vision inspection with AI is to have enough quality data. However, it is not easy to build a system for acquiring manufacturing data, and it takes more time to collect data with unspecified defects. In this paper, we propose a data augmentation vision inspection technique using generative AI-based manufacturing data. In order to acquire high-quality data, we augmented the data by finding the optimal parameter combination with the input data of Stable Diffusion, and executed vision inspection based on it, and the results showed excellent accuracy. This approach is expected to overcome the difficulty of data collection and contribute to the improvement of product quality.
참고문헌 (16)
- Kim, H, "Development of an Effective Defect Classification System for Inspection of QFN Semiconductor Packages", The Journal of Korea Institute of Convergence Signal Processing, 10(2), pp.120~126 (2009)
- Park, Y., "Development of vision system for quality inspection of automotive parts and comparison of machine learning models", The Journal of the Convergence on Culture Technology (2022)
- Kim, H., "Exterior vision inspection method of injection molding automotive parts", Journal of the Korea Institute of Information and Communication Engineering (2019)
- Han, M. Y., "Development of Vision System for Quality Inspection of Machined Holes of Automobile Mechanical" (2023)
- Goodfellow, I., "Generative adversarial nets", Advances in Neural Information Processing Systems, 27, pp.2672~2680 (2014)
- Choi, Y., "StarGAN: Unified generative adversarial networks for multi-domain image-to-image translation", Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition, pp.8789~8797 (2018)
- Kim, S., "WGAN based time-series data augmentation for resolving data imbalance", ~ (2023)
- Cho, E., "Augmentation of Defect Image Data using GAN", ~ (2022)
- Cho, J. -H., "Super resolution data augmentation method for improve DCGAN image quality deterioration problems", Journal of the Korea Institute of Information and Communication Engineering, 28(3), pp.260~266 (2024)
- Ho, J., "Denoising diffusion probabilistic models", Advances in Neural Information Processing Systems, 33, pp.6840~6851 (2020)
- SiO, J, "Data Augmentation of Shelf Product for Object Recognition in O2O Stores Based on Generative AI", ~ (2024)
- Kim, S., "Design of a Waste Generation Model based on the Chat-GPT and Diffusion Model for data balance", ~ (2023)
- Lee, J. -H., "Image data augmentation technique combining diffusion and layout methods", Proceedings of Dong-Eui University, pp.1~10 (2024)
- Shim, S. -B., "A study on generation quality comparison of concrete damage image using stable diffusion base models", Journal of the Korea Institute for Structural Maintenance and Inspection, 28(4), pp.55~61 (2024)
- Rombach, R., "High-Resolution Image Synthesis with Latent Diffusion Models", Proceedings of the IEEE, 2022, 10684~10695
- He, K., "Deep residual learning for image recognition", Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition, pp.770~778 (2016)