Enhancing Glass Defect Detection with Diffusion Models: Addressing Imbalanced Datasets in Manufacturing Quality Control
Synthetic defect images from a diffusion model improve rare-defect recall across three vision architectures while preserving strong precision for automated glass quality inspection.
Rare defects make manufacturing inspection datasets difficult to balance. This work uses a denoising diffusion probabilistic model to generate additional defective-glass examples for training machine-vision classifiers.
Across ResNet50V2, EfficientNetB0, and MobileNetV2, the augmented data improved recall while maintaining perfect validation precision. ResNet50V2 showed the largest reported gain, increasing from 78 percent to 93 percent recall.
Have a real process in mind?
Turn operational friction into a practical AI project.
Tell us where work is slow, repetitive, or hard to scale. We’ll help you find the smallest useful place to start.