Search papers, labs, and topics across Lattice.
This paper introduces Wireframe Detection and Repair (WDR), a training-free framework designed to enhance the validity of boundary representation (B-Rep) generation by addressing geometric and topological anomalies in intermediate wireframes. The method employs a Geometric-Topology Anomaly Detector (GTAD) that uses a combination of visual language models and specialized detectors to assess and mitigate risks of invalid B-Reps without the need for retraining large generative models. Experimental results indicate that WDR significantly improves the validity of generated CAD models while maintaining their diversity and quality, showcasing its potential for practical applications in CAD synthesis pipelines.
Training-free anomaly detection and repair can drastically enhance the validity of CAD models without the computational burden of retraining large generative models.
Multi-stage boundary representation (B-Rep) generation leverages intermediate wireframes to synthesize CAD models. However, geometric and topological risks in these wireframes -- such as self-intersections, edge collapses, and disconnected vertices -- can propagate to invalid final B-Reps. Mitigating such failures by retraining large generative models is computationally prohibitive. We propose Wireframe Detection and Repair (WDR), a training-free framework that intervenes at the intermediate wireframe stage to improve downstream B-Rep validity. WDR features a Geometric-Topology Anomaly Detector (GTAD) that combines parallel VLM-based coarse screening with geometric and topological detectors to predict downstream invalidity risk and route generation to dedicated branches. An Energy-Guided Geometric-Topology Repair (EGGTR) module then performs detector-triggered guided regeneration through geometry and topology branches. By scaling test-time computation via Energy-Guided Resampling and training-free guidance for diffusion models, WDR can be integrated into autoregressive and diffusion pipelines without retraining. Extensive experiments demonstrate consistent improvements in kernel-checked validity while largely retaining the measured diversity and distributional quality of synthesized CAD models. The code will be made publicly available upon acceptance.
