Casting Defect Detection: 12-Class Synthetic Images for Industrial Inspection

Assembled from synthetic images for casting defect detection, categorized into 12 distinct defect classes. It is designed for computer vision applications in industrial quality control and inspection.

Use Cases

• Train a multi-class image classifier to identify the 12 defect categories for automated visual inspection.
• Benchmark object detection models on synthetic casting defect images to localize flaws in industrial parts.
• Use the 12-class labeled synthetic data to augment real-world inspection datasets and improve model generalization.

Strengths

• Contains images for 12 distinct defect classes, providing a structured multi-class problem.
• Synthetic data generation allows for controlled variation and potentially large-scale dataset creation.

Limitations

• Synthetic images may not fully capture the texture, lighting, and noise variations of real-world casting environments.
• The dataset's size, resolution, and diversity of synthetic defects are unknown, limiting performance assessment.

Provenance

Source

Kaggle

Collection Method

Synthetically generated images for industrial computer vision.