MFCAD-VLM: Manufacturing CAD Feature Dataset with STEP Files, Ground Truth JSON Annotations, and Multi-View Isometric Images for Automatic Feature Recognition

This dataset is designed to advance research in Computer-Aided Design (CAD) and Automatic Feature Recognition (AFR) by providing a comprehensive set of files supporting a range of manufacturing processes, including machining, additive manufacturing, sheet metal forming, molding, and casting. The dataset includes three main folders:

1. STEP Files: Contains CAD models in STEP format, representing a variety of parts with distinct manufacturing features. These designs are categorized by complexity levels (easy, medium, and hard) based on feature type and quantity.

2. Ground Truth JSON Files: Each JSON file aligns with a corresponding STEP file and contains detailed annotations by experts. These annotations define manufacturing feature types, quantities, and additional specifics essential for accurate AFR assessment.

3. Multi-View Isometric Images: For each CAD model, this folder provides three isometric-view snapshots generated via Python. These images capture the model from different viewing angles, offering diverse visual perspectives to aid feature recognition tasks.

This dataset serves as a resource for evaluating and benchmarking AFR models, particularly those using vision-language models and prompt engineering. Researchers are encouraged to refer to the associated paper for methodology, dataset structure, and additional details regarding image generation and feature categorization.