Shulou(Shulou.com)11/24 Report--

In the whole house customization industry scene, CAD drawing is an important way to present the design plan. However, the CAD drawings completed by the store designer can not be directly identified by the machine, so it is often necessary to manually transform the design language contained in the CAD drawings into a three-dimensional model, and then further generate data that can be processed by the production line, which has a great impact on data accuracy, order removal efficiency and factory production capacity.

In order to solve this problem, group core technology has been focusing on the research of 3D modeling using AI in recent years. Recently, a paper entitled "PlankAssembly: Robust 3D Reconstruction from Three Orthographic Views with Learnt Shape Programs" was accepted by the International computer Vision Conference (ICCV) in cooperation with Sun Yat-sen University, University of Electronic Science and Technology and other academic institutions. ECCV, ICCV and CVPR are known as the three top international academic conferences in the field of computer vision.

This paper studies a long-standing problem in computer-aided design: the reconstruction of 3D model from orthographic views. The traditional algorithm solves this problem by establishing the explicit relationship between two-dimensional drawings and three-dimensional models. Experiments show that the traditional algorithm is very sensitive to drawing errors, which limits the application of this kind of algorithm in practical products. At present, some colleagues try to identify 2D information such as length and width by means of geometric recognition, but they can only achieve simple CAD model reconstruction.

In order to solve the problem that the algorithm is sensitive to drawing errors, Group Core Technology tries to use AI deep learning algorithm based on data-driven means to alleviate this problem. Specifically, this study uses the self-attention mechanism for implicit modeling. By training in the data set containing more than 26000 parameterized CAD models, the 2D line drawings in three orthographic views can be automatically converted into 3D CAD models. Compared with the traditional algorithm, the robustness of our proposed algorithm to noise is significantly improved.

More importantly, generally speaking, the reconstructed objects can not be edited, but this study reconstructs the linkage relationship between the panels in order to achieve the editable effect on the reconstructed cabinet.

Based on the algorithm research proposed by group nuclear science and technology, this research provides a new idea of 3D modeling for the industry, and this research is supported by the key R & D project of Zhejiang Province in 2022.

An overseas practitioner in the artificial intelligence industry was surprised by the results of the study and described it as a carpenter and a sharp weapon for 3D printers.

In order to solve the problem of CAD reconstruction of customized cabinets more quickly, Group Nuclear Technology will accelerate the application of the results of this research in industrial scenarios.

Welcome to subscribe "Shulou Technology Information " to get latest news, interesting things and hot topics in the IT industry, and controls the hottest and latest Internet news, technology news and IT industry trends.