Self-supervised Graph Neural Network for Mechanical CAD Retrieval

Yuhan Quan,Huan Zhao,Jinfeng Yi,Yuqiang Chen

2024-06-18

Abstract:CAD (Computer-Aided Design) plays a crucial role in mechanical industry, where large numbers of similar-shaped CAD parts are often created. Efficiently reusing these parts is key to reducing design and production costs for enterprises. Retrieval systems are vital for achieving CAD reuse, but the complex shapes of CAD models are difficult to accurately describe using text or keywords, making traditional retrieval methods ineffective. While existing representation learning approaches have been developed for CAD, manually labeling similar samples in these methods is expensive. Additionally, CAD models' unique parameterized data structure presents challenges for applying existing 3D shape representation learning techniques directly. In this work, we propose GC-CAD, a self-supervised contrastive graph neural network-based method for mechanical CAD retrieval that directly models parameterized CAD raw files. GC-CAD consists of two key modules: structure-aware representation learning and contrastive graph learning framework. The method leverages graph neural networks to extract both geometric and topological information from CAD models, generating feature representations. We then introduce a simple yet effective contrastive graph learning framework approach, enabling the model to train without manual labels and generate retrieval-ready representations. Experimental results on four datasets including human evaluation demonstrate that the proposed method achieves significant accuracy improvements and up to 100 times efficiency improvement over the baseline methods.

Information Retrieval,Artificial Intelligence,Computer Vision and Pattern Recognition

What problem does this paper attempt to address?

### Problems Addressed by the Paper

The paper "Self-supervised Graph Neural Network for Mechanical CAD Retrieval" aims to address the issue of efficient reuse of mechanical CAD (Computer-Aided Design) components. Specifically, the paper focuses on the following two key challenges:

1. **Limitations of Feature Extraction**: Existing CAD retrieval methods, whether based on text or relying on 3D data descriptors, often fail to fully capture the rich geometric and topological information in CAD models. This results in poor generalization ability and high computational costs.

2. **High Annotation Cost**: Although learning-based methods have good generalization ability, one major problem in training an effective model is the extremely high cost of annotating CAD similarity labels. Unlike text documents, annotating the similarity of CAD components requires deeper mechanical engineering knowledge, which significantly increases the cost. Therefore, unsupervised learning methods are crucial for making CAD retrieval practical and scalable.

To address these challenges, the paper proposes GC-CAD (Graph Contrastive learning for CAD retrieval), a novel self-supervised graph learning framework. GC-CAD directly models from the BRep format of CAD components and utilizes graph neural networks for representation learning. By adopting an effective contrastive learning approach, GC-CAD can be trained using unlabeled CAD data, avoiding the expensive annotation work. Experimental results show that GC-CAD significantly outperforms baseline methods on 4 datasets, including human evaluation results, with notable improvements in accuracy and efficiency.