TransCAD: A Hierarchical Transformer for CAD Sequence Inference from Point Clouds
This addresses the problem of automating CAD model reconstruction from 3D scans for applications in manufacturing and design, representing a strong domain-specific advancement.
The paper tackles 3D reverse engineering by proposing TransCAD, a hierarchical transformer architecture that predicts CAD sequences from point clouds, achieving state-of-the-art results on DeepCAD and Fusion360 datasets with a new metric called mean Average Precision of CAD Sequence.
3D reverse engineering, in which a CAD model is inferred given a 3D scan of a physical object, is a research direction that offers many promising practical applications. This paper proposes TransCAD, an end-to-end transformer-based architecture that predicts the CAD sequence from a point cloud. TransCAD leverages the structure of CAD sequences by using a hierarchical learning strategy. A loop refiner is also introduced to regress sketch primitive parameters. Rigorous experimentation on the DeepCAD and Fusion360 datasets show that TransCAD achieves state-of-the-art results. The result analysis is supported with a proposed metric for CAD sequence, the mean Average Precision of CAD Sequence, that addresses the limitations of existing metrics.