基于简约线条图的三维模型Z轴切片视图生成数据

在3D内容生成和打印应用中，使用单一线条简图生成Z轴方向的多层切片图像。此方法能够将设计者创作的草图或概念图转化为沿Z轴的多层次三维视图切片，特别适用于呈现复杂的结构和遮挡的细节区域，有效提高了3D模型创建的效率和准确度。该技术不仅对工业设计和艺术品制作有帮助，还能够在教育和研究中便捷生成3D内容，极大提升了用户的创作效率。1. 数据收集：在3D设计和3D打印应用场景中，获取设计者制作的单一视角线条图，并生成与之对应的三维模型切片。利用公司内部的数据采集流程，从单张线条图提取相关的三维结构数据和在Z轴方向上的切片图像。每个三维模型包含三维渲染图、简约线条图。
2. 数据处理：在预处理步骤中，将三维模型沿Z轴方向进行多角度切割，得到Z轴切片渲染图和Z轴切片简约线条图。通过图像分析技术与点云转换，提取线条图的特征向量，并使用公式F_sketch = Encoder_sketch(Sketch_img)进行特征编码。此处的Encoder_sketch是预先训练好的线条图编码器，Sketch_img则为输入的简约线条图。Z轴切片简约线条图的特征编码通过另一编码器Encoder_slice生成切片特征F_slice。最终将F_sketch与F_slice匹配，以保持Z轴切片简约线条图与简约线条图在特征上的一致性。
3. 模型构建：设计了一个基于自编码器的神经网络，该模型以简约线条图特征为输入并生成Z轴切片简约线条图。通过公式Slice_gen = Decoder_slice(F_sketch)生成切片预测结果，其中Decoder_slice是用于解码三维切片图的模块。最后以平均Acc指标评估模型性能。

In 3D content generation and printing applications, single simplified line sketches are used to generate multi-layer sliced images along the Z-axis. This method can convert sketches or conceptual drawings created by designers into multi-layered 3D view slices along the Z-axis, which is particularly suitable for presenting complex structures and occluded detail areas, effectively improving the efficiency and accuracy of 3D model creation. This technology not only assists industrial design and art production, but also facilitates the generation of 3D content in education and research, greatly enhancing users' creative efficiency.

1. Data Collection: In 3D design and 3D printing application scenarios, single-view line drawings created by designers are collected, and corresponding 3D model slices are generated. Using the company's internal data collection process, relevant 3D structure data and sliced images along the Z-axis are extracted from a single line drawing. Each 3D model includes 3D renderings and simplified line sketches.

2. Data Processing: In the preprocessing step, 3D models are cut at multiple angles along the Z-axis to obtain Z-axis sliced renderings and Z-axis sliced simplified line sketches. Feature vectors of the line drawings are extracted through image analysis technology and point cloud transformation, and feature encoding is performed using the formula $F_{ ext{sketch}} = ext{Encoder}_{ ext{sketch}}( ext{Sketch_img})$. Here, $ ext{Encoder}_{ ext{sketch}}$ is a pre-trained line drawing encoder, and $ ext{Sketch_img}$ is the input simplified line sketch. The feature encoding of the Z-axis sliced simplified line sketches generates slice features $F_{ ext{slice}}$ through another encoder $ ext{Encoder}_{ ext{slice}}$. Finally, $F_{ ext{sketch}}$ and $F_{ ext{slice}}$ are matched to maintain feature consistency between the Z-axis sliced simplified line sketches and the input simplified line sketches.

3. Model Construction: An autoencoder-based neural network is designed, which takes the features of simplified line sketches as input and generates Z-axis sliced simplified line sketches. Slice prediction results are generated using the formula $ ext{Slice\_gen} = ext{Decoder}_{ ext{slice}}(F_{ ext{sketch}})$, where $ ext{Decoder}_{ ext{slice}}$ is the module used for decoding 3D slice images. Finally, the model performance is evaluated using the average Acc metric.

搜集汇总

数据集介绍

特点

该数据集包含7129条基于简约线条图生成三维模型Z轴切片视图的数据，适用于3D内容生成和打印应用，通过特定算法提升模型创建效率和准确度。