Drawing2CAD: Sequence-to-Sequence Learning for CAD Generation from Vector Drawings
Feiwei Qin, Shichao Lu, Junhao Hou, Changmiao Wang, Meie Fang, Ligang Liu
2025-09-05
Summary
This research introduces a new system called Drawing2CAD that automatically creates 3D CAD models from standard 2D engineering drawings, like blueprints. It's a way to bridge the gap between how designs are traditionally started and the latest advances in AI-powered design tools.
What's the problem?
Currently, a lot of progress has been made in creating 3D models from things like images or text, but these methods don't really fit with how engineers actually work. Engineers usually start with 2D drawings, and there hasn't been much work on automatically turning those drawings into the precise, editable 3D CAD models they need. Existing methods don't preserve the exactness and original design ideas from the 2D drawings.
What's the solution?
The researchers treated the problem like translating one language to another – specifically, translating the 'language' of 2D drawings into the 'language' of 3D CAD commands. They developed a system that breaks down drawings into basic shapes and then uses a special type of AI network, called a transformer, to figure out which 3D commands to use and how to adjust them. They also created a new dataset of paired drawings and CAD models to train and test their system, and focused on making sure the system understands the flexibility inherent in CAD design.
Why it matters?
This is important because it could significantly speed up the design process for engineers. Instead of manually recreating a 3D model from a 2D drawing, the system can do it automatically, saving time and reducing the chance of errors. It also allows for a more seamless integration of AI into existing engineering workflows, making design more efficient and potentially opening up new possibilities for innovation.
Abstract
Computer-Aided Design (CAD) generative modeling is driving significant innovations across industrial applications. Recent works have shown remarkable progress in creating solid models from various inputs such as point clouds, meshes, and text descriptions. However, these methods fundamentally diverge from traditional industrial workflows that begin with 2D engineering drawings. The automatic generation of parametric CAD models from these 2D vector drawings remains underexplored despite being a critical step in engineering design. To address this gap, our key insight is to reframe CAD generation as a sequence-to-sequence learning problem where vector drawing primitives directly inform the generation of parametric CAD operations, preserving geometric precision and design intent throughout the transformation process. We propose Drawing2CAD, a framework with three key technical components: a network-friendly vector primitive representation that preserves precise geometric information, a dual-decoder transformer architecture that decouples command type and parameter generation while maintaining precise correspondence, and a soft target distribution loss function accommodating inherent flexibility in CAD parameters. To train and evaluate Drawing2CAD, we create CAD-VGDrawing, a dataset of paired engineering drawings and parametric CAD models, and conduct thorough experiments to demonstrate the effectiveness of our method. Code and dataset are available at https://github.com/lllssc/Drawing2CAD.