Generates validated synthetic defect images to expand training datasets where real defect samples are scarce, reducing dependence on costly or difficult-to-obtain physical defect samples and increasing defect-variation coverage.

Inspection and AI-training datasets may lack sufficient real defect images, particularly for rare or diverse defect types, limiting machine-learning model performance and the ability to adequately represent different defect variations.

Provides an AI-based synthetic-image-generation solution that categorises real images, creates validated defect masks, trains image-generation models using paired mask and real-image datasets, and generates synthetic images that are validated and refined through similarity and feedback mechanisms. This is closely aligned with MIMOS' current Mi-IndusData+ offering, which uses automated image synthesis and annotation to address industrial data scarcity for AI-based inspection.

Addresses limited real-world defect data by combining defect-mask generation, paired mask-image training, template creation, synthetic-image generation, similarity validation and feedback-based refinement to create usable AI-training datasets.