A Public Ground-Truth Dataset for Handwritten Circuit Diagram Images

<strong>CGHD</strong>This dataset contains images of hand-drawn electrical circuit diagrams as well as accompanying annotation and segmentation ground-truth files. It is intended to train (e.g. ANN) models for extracting electrical graphs from raster graphics.<br><strong>Content</strong><strong>2.549</strong> Raw Images (Annotated)25 Drafters (plus Images provided by TU Dresden)12 Circuits per Drafter2 Drawings per Circuit4 Photos per Drawing<strong>212.280 </strong>Bounding Box Annotations<strong>29.790</strong> Rotation Annotations<strong>71.307</strong> Text String Annotations<strong>264</strong> Binary Segmentation Maps (Annotated)Strokes vs. BackgroundAccompanying Polygon Annotation Files<strong>20.060</strong> Polygon Annotations<strong>59 </strong>Object Classes<strong>Scripts</strong> for Data Loading, Statistics, Consistency Check and Training Preparation

<strong>CGHD数据集</strong>本数据集包含手绘电气电路图图像，以及配套的标注文件与分割真值文件，旨在训练（例如人工神经网络(Artificial Neural Network, ANN)）模型，以从光栅图像中提取电气拓扑图。<br><strong>数据集内容</strong><br><strong>2,549</strong> 张带标注的原始图像；共25名绘图者（另有德累斯顿工业大学(TU Dresden)提供的图像），每名绘图者对应12种电路，每种电路包含2幅图纸，每幅图纸配有4张拍摄照片。<br><strong>212,280</strong> 个边界框标注<br><strong>29,790</strong> 个旋转标注<br><strong>71,307</strong> 个文本字符串标注<br><strong>264</strong> 张带标注的二值分割掩码，用于区分绘图笔画与背景<br>配套多边形标注文件：<br><strong>20,060</strong> 个多边形标注<br><strong>59</strong> 个对象类别<br><strong>脚本</strong>：附带用于数据加载、统计分析、一致性校验与训练预处理的脚本文件。

搜集汇总

背景与挑战

背景概述

该数据集包含手绘电路图的图像及相应的标注和分割真实数据，旨在训练模型从栅格图形中提取电气图。数据集包含2,549张原始图像、212,280个边界框标注和59个对象类别，适用于人工智能和图像处理研究。