CAD Classification and Multi-Labeling Using HOOPS Embeddings

These notebooks demonstrate two ways to assign categories to CAD parts using shape embeddings: aggregate labels from similar parts without retraining, or train a supervised classifier on pre-computed embeddings.

Follow Install HOOPS AI for environment and license setup. Both notebooks reuse the index built in CAD Indexing and Retrieval Using HOOPS Embeddings. The Predicting CAD Metadata with the HOOPS AI Context Layer quickstart introduces the metadata prediction workflow used by the multi-labeling notebook.

Choose the approach that fits your labeling workflow:

1. demo_HOOPS_Embeddings_Multi-labeling.ipynb combines similarity search with a custom Context Layer aggregation rule to predict several categories per part without model retraining.

2. demo_HOOPS_Embeddings_MLP_classifier.ipynb trains and evaluates an MLP classifier on existing embeddings, calibrates confidence scores, and shows how to adapt single-label classification to a multi-label problem.

The code and resources for this tutorial can be found in the HOOPS-AI-Tutorials GitHub repository.

Info

From the repository root, run jupyter lab embeddings_pipeline/demo_HOOPS_Embeddings_Multi-labeling.ipynb for the retrieval-based approach, or jupyter lab embeddings_pipeline/demo_HOOPS_Embeddings_MLP_classifier.ipynb to train a classifier.

• Predicting multiple categories with the HOOPS AI Context Layer
• Configure
• Classifying CAD parts by training an MLP on shape embeddings