TreasureX Researchcs.AIFake Defects, Real Savings: AI-Generated Training Data for Printing Quality Control
Korota Arsène Coulibaly, Mohamed Hamlich, Khalid Hmali, Andrea Trombin · July 23, 2026
In the world of gravure printing—used for magazines, packaging, and wallpaper—quality control still relies on slow, costly manual inspection. Deep learning models could automate defect detection, but they need thousands of real defect images, which are extremely rare in industry. This paper presents a synthetic data generation framework that creates realistic images of printing defects (creases, streaks, misregistration) along with their bounding-box annotations. Using 7,533 synthetic images to train the RFDETR object-detection model, the system achieved 80.9% mean average precision on real industrial test samples. The framework is zero-cost and can be rapidly deployed, enabling printing lines to automate inspection without massive manual data collection. This approach could slash inspection costs, speed up production, and improve consistency—a practical win for manufacturers and a glimpse into how synthetic data can solve real-world data scarcity problems.
Read more: https://arxiv.org/abs/2607.21577
Educational content only, generated with AI assistance from public research abstracts. Not investment advice. Thank you to arXiv for use of its open access interoperability.