GeoBiked
收藏资源简介:
GeoBiked数据集由宝马集团和奥格斯堡大学创建,包含4355张自行车图像,这些图像带有结构和技术特征的注释。数据集旨在支持工程设计中的深度生成模型(DGMs),通过自动化标签技术提高数据标注效率。数据集内容包括19种不同自行车风格的分类、前轮和后轮直径、车架和管尺寸等技术特征,以及12个几何参考点的坐标。创建过程包括从BIKED数据集的4512张图像中筛选和标准化,确保几何一致性。该数据集主要应用于工程设计领域,旨在解决数据驱动设计中的数据稀缺问题,支持模型训练、微调和条件机制的开发。
The GeoBiked dataset, developed by BMW Group and the University of Augsburg, consists of 4355 bicycle images annotated with structural and technical characteristics. It is designed to support deep generative models (DGMs) in engineering design, with the goal of enhancing data annotation efficiency through automated labeling technologies. The dataset covers annotations for 19 distinct bicycle style categories, technical features including front and rear wheel diameters, frame and tube dimensions, as well as coordinates of 12 geometric reference points. Its construction involved screening and standardization from 4512 images sourced from the BIKED dataset to ensure geometric consistency. Primarily applied in the engineering design domain, this dataset addresses the data scarcity problem in data-driven design, and supports model training, fine-tuning, and the development of conditional mechanisms.

- 1GeoBiked: A Dataset with Geometric Features and Automated Labeling Techniques to Enable Deep Generative Models in Engineering Design宝马集团,奥格斯堡大学 · 2024年



