Voxel51/DeepPatent

DeepPatent是一个大规模的技术绘图数据集，从美国设计专利文档中提取。该数据集包含了按专利组（基于发布日期）和组织结构排列的专利绘图，每个专利包含多个绘图图像，展示了专利设计的不同视角和方面。该数据集使用FiftyOne分组数据集结构进行组织，每个专利作为一个组，包含多个绘图片段。数据集由Michal Kucer、Diane Oyen、Juan Castorena和Jian Wu策划，语言为英语，采用BSD-3许可证。

DeepPatent is a large-scale dataset of technical drawings extracted from U.S. design patent documents. The dataset contains patent drawings organized by patent groups (based on publication dates) and individual patent numbers, with each patent containing multiple drawing images showing different views and aspects of the patented designs. This implementation organizes the data into a FiftyOne grouped dataset structure, where each patent serves as a group containing multiple drawing slices. The dataset is curated by Michal Kucer, Diane Oyen, Juan Castorena, and Jian Wu, with the language being English and is licensed under the BSD-3 License.