introvoyz041/sam-dataset

该数据集包含超过600万个从Onshape提取的CAD 2D草图。草图以自定义SAM格式的Python对象存储，利用Sketchgraphs数据集满足工业需求，并支持在其他CAD软件上更轻松的迁移学习。支持的任务包括自动草图生成和自动约束。数据实例以npy文件形式存储，包含Python pickled对象，需使用Sketchgraphs的flat_array模块加载。数据分割为训练集600万样本，验证集和测试集各5万样本。This dataset contains over 6M CAD 2D sketches extracted from Onshape. Sketches are stored as python objects in the custom SAM format. SAM leverages the Sketchgraphs dataset for industrial needs and allows for easier transfer learning on other CAD softwares. Supported tasks include Automatic Sketch Generation and Auto Constraint. Data instances are stored in npy files containing python pickled objects, requiring the flat_array module from Sketchgraphs for loading. Data splits consist of 6M training samples, 50k validation samples, and 50k test samples.