⌁ Awesome AI4CAD Hub Papers · Datasets · Projects
中文 GitHub
← Back to papers

CMT: A Cascade MAR with Topology Predictor for Multimodal Conditional CAD Generation

arXiv 2025 59 method, benchmark

TLDR

A cascade MAR with topology predictor for multimodal B-Rep CAD generation, plus a large-scale dataset mmABC with 1.3M models.

Reasoning

The paper introduces a novel framework (CMT) for multimodal CAD generation using B-Rep, supported by a new large-scale dataset mmABC. Strengths include being the first multimodal B-Rep framework and strong empirical results; weaknesses include lack of discussion on limitations and potential over-reliance on the dataset.

Read-first score

Read-first score 59, weighted from topical fit, citation, graph, method, reproducibility, and recency signals. Original total remains 97.

Recency 8%
86.7

Uses a gentle age decay so recent papers surface without erasing older foundations. 2025

Topical relevance 42%
60.6

Uses existing LLM keyword relevance scores normalized to 0-100. AI for CAD,computer-aided design,neural CAD,generative CAD,parametric CAD,B-Rep,boundary representation,constructive solid geometry,CSG,sketch extrusion,CAD generation,CAD reconstruction,text-to-CAD,image-to-CAD,point cloud to CAD,CAD program

Methodology quality 25%
60

Screens visible abstract and analysis fields for experiment, dataset, baseline, metric, and limitation evidence. markers=dataset,experiment

Reproducibility 25%
46

Screens links and visible text for paper, code, dataset, artifact, and repository signals. pdf=True; code=False; dataset=False; markers=code,dataset

Field roles

Frontier

Rank sensitivity

Stability: volatile; rank range: 22.

Keyword Scores

computer-aided design
10
B-Rep
10
boundary representation
10
CAD generation
10
generative CAD
9
AI for CAD
8
text-to-CAD
8
image-to-CAD
8
point cloud to CAD
8
neural CAD
7
CAD reconstruction
7
parametric CAD
2
constructive solid geometry
0
CSG
0
sketch extrusion
0
CAD program
0

Deep Analysis

Innovations

  • First multimodal framework for CAD generation based on Boundary Representation (B-Rep) using a cascade MAR with topology predictor (CMT)
  • Cascade MAR captures 'edge-counters-surface' priors essential in B-Reps
  • Topology predictor directly estimates topology in B-Reps from compact MAR tokens
  • Large-scale multimodal CAD dataset mmABC with over 1.3 million B-Rep models and multimodal annotations (point clouds, text, multi-view images)

Methodology

The paper proposes CMT, a cascade Masked Autoencoder (MAR) with a topology predictor for B-Rep-based CAD generation. The cascade MAR captures edge-counters-surface priors, while the topology predictor estimates topology from compact tokens. A new multimodal dataset mmABC with 1.3M B-Rep models and annotations (point clouds, text, multi-view images) is introduced for training.

Key Results

CMT improves Coverage by +10.68% and Valid ratio by +10.3% over state-of-the-art on ABC unconditional generation, and improves Chamfer distance by +4.01 on image-conditioned CAD generation on mmABC.

Resources

Paper

Tags