New Masked Topology Modeling enhances self-supervised learning for CAD data

Researchers have introduced Masked Topology Modeling (MTM), a novel self-supervised learning technique designed for parametric CAD data. MTM reconstructs a face-adjacency graph unique to boundary representations (B-reps) by predicting the convexity and curve type of masked edges. This method, combined with contrastive learning and B-rep-aware augmentations, demonstrates strong performance on various benchmarks, utilizing datasets like ABC and a new procedurally generated one. AI

IMPACT This new method could improve data efficiency in the design of modern objects by enabling better learning from limited CAD datasets.

RANK_REASON The cluster describes a new method presented in an academic paper on arXiv. [lever_c_demoted from research: ic=1 ai=1.0]

Read on arXiv cs.LG →

• ABC dataset
• arXiv
• B-reps
• Hugging Face
• Masked Topology Modeling

• paper
• other

AI-generated summary · Google Gemini · from 1 sources. How we write summaries →