Evaluates 3D CAD models against reference designs using boolean operations and geometric similarity metrics.
This skill provides a robust framework for comparing 3D models within the Claude Code environment, utilizing OpenCASCADE to perform precise boolean operations. It calculates industry-standard metrics such as Intersection over Union (IoU), Dice coefficients, and precision/recall scores to quantify the accuracy of generated geometry against gold references. Ideal for developers working on 3D generative AI or automated CAD validation, it generates detailed reports and GLB visualizations to identify specific areas of over-generation, under-generation, or positional offsets.
Key Features
01Supports high-fidelity CAD formats including STEP, BREP, and STL meshes.
02Calculates standard similarity metrics including IoU, Dice, Precision, and Recall for 3D volumes.
03Provides automated diagnostic interpretation of size, position, and volume discrepancies.
041 GitHub stars
05Generates GLB visualization files to isolate missing, extra, and common geometry.
06Includes a JSON-only output mode for seamless integration into ML training and evaluation pipelines.
Use Cases
01Evaluating the performance of AI-generated 3D models against ground-truth reference designs.
02Comparing revisions of mechanical CAD parts to track geometric changes and volume shifts.
03Implementing automated quality gates in 3D design workflows using standardized similarity metrics.
