A data generation method with dual discriminators and regularization for surface defect detection under limited data
Abstract: Highlights•Collecting surface defect images in industry presents significant challenges.•Limited defect data severely hampers the performance of defect detection models.•Our method can prevent GAN training crashes in situations where data is scarce.•Our novel data generation method produces high-quality and diverse defect images.•Our method enhances defect detection models’ performance in data-limited scenarios.
External IDs:dblp:journals/cii/LiuZYM23
Loading