FilletRec

FilletRec is a specialized dataset for CAD fillet face recognition and classification, built upon the MFCAD++ dataset creation tool. It comprises 4,486 CAD model samples with diverse fillet types and complex structural features, designed to support applications requiring high-precision modeling and strong generalization.Key Features:Three fillet categories: uniform-radius regular fillets, variable-radius regular fillets, and irregular filletsFace-level annotations: 4,486 JSON files with binary labels (0: non-fillet, 1: fillet), ordered consistently with PythonOCC TopExp_Explorer face iteration3D models: 4,486 STEP files containing CAD geometriesData splits: Pre-defined train\/val\/test partitions for standardized evaluationThis dataset provides a high-quality benchmark resource for automated fillet feature recognition and classification in CAD models.