Synthetic Data Generation Framework for Automated Quality Control in Rotogravure Printing (arXiv 2507.20473)

Forum topic · 小凯 · 2026-07-26

Summary

A new paper (arXiv:2507.20473) by Korota Arsene Coulibaly, Mohamed Hamlich, and Khalid Hmlich introduces a synthetic data generation framework for automated quality control in rotogravure printing. Manual inspection in printing is slow, costly, and subjective, while deep learning models such as YOLO and Vision Transformers are hindered by the extreme scarcity of real-world industrial defect images. The proposed pipeline automatically generates high-fidelity images of specific printing defects—such as creases, streaks, and misregistration—along with corresponding bounding boxes and annotations. The authors built a synthetic dataset of 7,533 images and used it to train the state-of-the-art object detection model RFDETR. The trained model achieved 80.9% mean average precision (mAP) on real industrial test samples, demonstrating strong sim-to-real generalization. The framework offers a low-cost, fast-to-deploy solution for automated defect inspection on printing lines without extensive manual data collection, addressing a key bottleneck in industrial computer vision applications.

Paper Overview

Field: Computer Vision
Authors: Korota Arsene Coulibaly, Mohamed Hamlich, Khalid Hmlich
Published: 2026-07-25
arXiv: 2507.20473

Abstract

Quality control in printing, particularly in rotogravure printing, still depends on slow, costly, and subjective manual inspection. Automated surface defect detection is critical for maintaining high-quality standards in rotogravure printing. Deep learning models give prospects for automation. However, training robust deep learning models, such as YOLO or Vision Transformers, is heavily hindered by the extreme scarcity of real-world industrial defect images.

To overcome this limitation, this paper introduces a novel synthetic data generation framework tailored for rotogravure printing quality control. The proposed pipeline automatically generates high-fidelity images of specific printing defects (creases, streaks, misregistration, etc.) and outputs corresponding bounding boxes and annotations.

Key Results

Generated a synthetic dataset of 7,533 images covering common printing defect types.
Trained the state-of-the-art object detection model RFDETR on the synthetic data.
Achieved 80.9% mAP on real industrial test samples, showing strong generalization from synthetic to real data.

Significance

The framework provides a zero-cost, fast-to-deploy solution for automated defect inspection on printing lines, eliminating the need for extensive manual data collection—a major bottleneck for deep learning in industrial quality control.

---

*Auto-collected on 2026-07-26*