BIKED++

BIKED++是由麻省理工学院创建的多模态数据集，包含140万自行车设计和参数化CAD图像。数据集通过BikeCAD软件生成，涵盖了从参数化设计到图像的转换过程。创建过程中，首先使用Sobol序列从设计空间中随机抽样，然后通过约束检查筛选出可行设计，并将其转换为BikeCAD文件，最后渲染成PNG图像。该数据集主要用于训练跨模态预测模型，解决参数化设计和图像设计之间的相似性问题，支持工程设计中的多模态学习。

BIKED++ is a multimodal dataset created by the Massachusetts Institute of Technology (MIT), which contains 1.4 million bicycle designs and parametric CAD images. Generated using BikeCAD software, this dataset covers the entire transformation process from parametric designs to images. During its creation, feasible designs are first randomly sampled from the design space via Sobol sequences, then screened through constraint checks, converted into BikeCAD files, and finally rendered into PNG images. This dataset is primarily used for training cross-modal prediction models, addressing the similarity issue between parametric designs and image-based designs, and supporting multimodal learning in engineering design.