MFCAD++ Dataset

Machining feature recognition dataset for deep learning consisting of B-Rep CAD models labelled on each B-Rep face with a machining feature. Dataset for paper: "Hierarchical CADNet: Learning from B-Reps for Machining Feature Recognition, Computer-Aided Design" - The MFCAD++ dataset is a machining feature recognition dataset containing B-Rep CAD models. - Each CAD model has been saved as a STEP file. - The CAD models were automatically generated using the PythonOCC CAD software. - For each CAD model, a machining feature class label is given on each B-Rep face. - These labels can easily be extracted from the STEP files. - The labels are given in "feature_labels.txt". - The dataset has been split into "train", "val" and "test" directories using a 70:15:15 split as per the original paper. - These splits are given in the files: "train.txt", "val.txt" and "test.txt". - There are training_set=41766, val_set=8950 & test_set=8949 samples, with 59665 samples in total. - For more information on hierachical B-Rep graphs store in H5DF files see "h5_structure.h5". If used please reference the paper: Colligan AR, Robinson TR, Nolan DC, Hua Y, Cao W. Hierarchical CADNet: Learning from B-Reps for Machining Feature Recognition, Computer-Aided Design

本数据集为面向深度学习的加工特征识别数据集，由所有B-Rep面均标注有加工特征的边界表示（B-Rep）计算机辅助设计（CAD）模型构成，相关研究成果发表于期刊《Computer-Aided Design》的论文《Hierarchical CADNet: Learning from B-Reps for Machining Feature Recognition》。
- MFCAD++ 数据集是一款包含B-Rep CAD模型的加工特征识别数据集。
- 每份CAD模型均以STEP文件格式存储。
- 该批CAD模型通过PythonOCC CAD软件自动生成。
- 针对每份CAD模型，其每一个B-Rep面均配有对应的加工特征类别标签。
- 此类标签可直接从STEP文件中便捷提取。
- 所有标签信息均存储于"feature_labels.txt"文件中。
- 数据集按照原论文设定的70:15:15的比例，划分为训练集（train）、验证集（val）与测试集（test）三个目录。
- 数据集的划分信息分别存储于"train.txt"、"val.txt"与"test.txt"三个文件中。
- 数据集总计包含59665个样本，其中训练集41766个、验证集8950个、测试集8949个。
- 若需了解存储于H5DF文件中的分层B-Rep图的更多细节，请参考"h5_structure.h5"文件。若使用本数据集，请引用以下论文：Colligan AR, Robinson TR, Nolan DC, Hua Y, Cao W. Hierarchical CADNet: Learning from B-Reps for Machining Feature Recognition, Computer-Aided Design

搜集汇总

背景与挑战

背景概述

MFCAD++数据集是一个用于深度学习的加工特征识别数据集，包含基于B-Rep表示的CAD模型，每个模型的B-Rep面都标注了加工特征类别标签，便于机器学习任务。数据集以STEP文件格式提供，包含59,665个样本，已按70:15:15的比例分割为训练、验证和测试集，适用于计算机辅助设计领域的特征识别研究。