基于二维简约线条图的CAD模型数据

通过图像处理算法合成的二维简约线条图数据，利用深度生成模型进行训练，算法可以根据用户绘制的线条生成相应的CAD模型数据。这一数据适用于工业设计与制造领域，用户能够通过简单的二维线条绘制快速生成精准的CAD模型，从而提升设计效率和准确性，解决传统CAD建模过程中耗时长、操作复杂的问题。在移动触控设备上绘制线条图，通过算法解析并生成对应的CAD指令，从而加速三维建模过程，可用于工业设计与制造领域。具体过程如下：（1）数据收集：使用相关渲染软件在随机相机视点下渲染出三维模型的RGB投影图。（2）数据处理：使用Canny等算法得到二维简约线条合成图。再使用图像编码器提取特征。公式为：F_lines = Encoder_lines(Image_lines)，其中，F_lines表示二维简约线条图特征向量，Encoder_lines表示预训练的图像编码器，Image_lines表示二维简约线条合成图。（3）模型构建：使用二维简约线条图特征向量作为自变量，CAD指令为因变量，设计并搭建深度学习模型。通过公式提取CAD指令：CAD_commands = Decoder_lines(F_lines)，其中，CAD_commands表示CAD指令，Decoder_lines为深度学习模型。然后将生成的CAD指令解析成三维模型。最后使用平均FID和平均IoU评估整体生成模型的质量。

2D simplified line drawing data synthesized via image processing algorithms. This dataset is used to train an algorithm based on deep generative models, which can generate corresponding CAD model data based on lines drawn by users. This dataset is applicable to industrial design and manufacturing scenarios, allowing users to rapidly generate accurate CAD models via simple 2D line drawings, thereby improving design efficiency and accuracy, and addressing the issues of time-consuming and overly complex operations in traditional CAD modeling workflows. Drawing line diagrams on mobile touch-enabled devices, followed by algorithmic parsing and generation of corresponding CAD instructions, can accelerate 3D modeling processes, and is suitable for industrial design and manufacturing applications. The specific process is as follows:
(1) Data collection: Use relevant rendering software to render RGB projection images of 3D models from random camera viewpoints.
(2) Data processing: Use algorithms such as the Canny edge detector to generate 2D simplified line composite images. Then extract features using an image encoder. The formula is: F_lines = Encoder_lines(Image_lines), where F_lines represents the feature vector of the 2D simplified line drawing, Encoder_lines represents the pre-trained image encoder, and Image_lines represents the synthesized 2D simplified line composite image.
(3) Model construction: Take the feature vectors of the 2D simplified line drawings as independent variables and CAD commands as dependent variables to design and build a deep learning model. Extract CAD commands via the formula: CAD_commands = Decoder_lines(F_lines), where CAD_commands represents the generated CAD instructions, and Decoder_lines is the deep learning model. Then parse the generated CAD commands into a 3D model. Finally, use average FID and average IoU to evaluate the quality of the overall generative model.

搜集汇总

数据集介绍

特点

该数据集包含56557条基于二维简约线条图的CAD模型数据，适用于工业设计与制造领域，通过深度生成模型训练，能够根据用户绘制的线条快速生成精准的CAD模型，提升设计效率和准确性。