Data supporting figures and tables in Attention-Based Framework for Automated Symbol Recognition and Wiring Design in Electrical Diagrams

<b>Dataset Description for Figshare</b>This dataset supports the findings of the research titled:<b>"Attention-Based Framework for Automated Symbol Recognition and Wiring Design in Electrical Diagrams"</b><b>Authors:</b> Ikenna Ekeke, Carlos Francisco Moreno-García, Eyad Elyan<br><b>Affiliation:</b> Robert Gordon University, Aberdeen, Scotland, UK<br><b>Date:</b> April 2025<b>Overview</b>The research presents an end-to-end deep learning framework combining <b>YOLOv8 object detection</b> with <b>attention mechanisms</b> to improve symbol recognition in electrical diagrams, followed by a <b>graph-based wiring algorithm</b> that automates wire routing between detected symbols. The system is tested across <b>proprietary and public datasets,</b> including:CGHD (Circuit Graph Hand-drawn Diagrams)DCD (Digital Circuit Diagrams)<b>Datasets Included</b>This data bundle includes the <b>primary data</b> used to generate figures and tables within the manuscript:Model performance metrics across different attention modules (Table 1 - Table 3).Class distribution tables (used in Figures 11a, 11b, 12a, 12b).Class-wise accuracy table (Table 4).Results of statistical analysis (Table 5, Duncan’s test).Validation Results on CGHD and DCD_Datasets (Table 6).Wiring algorithm evaluation metrics (Table 7).Each table is provided in <b>CSV format</b> that maps the data file to the corresponding figure or table in the paper.<b>Use Cases</b>This dataset is useful for:Benchmarking attention-based models for electrical symbol recognition.Studying the impact of class imbalance on detection accuracy.Reproducing automated wiring design evaluations using pathfinding algorithms.

<b>Figshare数据集描述</b>
本数据集支持题为<b>《基于注意力机制的电气图自动符号识别与布线设计框架》</b>的研究成果。
<b>作者：</b>Ikenna Ekeke、Carlos Francisco Moreno-García、Eyad Elyan<br>
<b>所属机构：</b>英国苏格兰阿伯丁罗伯特戈登大学<br>
<b>日期：</b>2025年4月

<b>概述</b>
本研究提出一种端到端深度学习框架，将YOLOv8目标检测与注意力机制（attention mechanisms）相结合，以提升电气图中的符号识别性能，随后采用基于图的布线算法实现检测符号间的自动布线。该系统在专有和公开数据集上进行了测试，包括：CGHD（Circuit Graph Hand-drawn Diagrams）、DCD（Digital Circuit Diagrams）。

<b>包含的数据集</b>
本数据包包含用于生成手稿中图表的原始数据：模型性能指标（不同注意力模块，表1-表3）、类别分布表（用于图11a、11b、12a、12b）、类别级准确率表（表4）、统计分析结果（表5，Duncan’s test）、CGHD和DCD数据集的验证结果（表6）、布线算法评估指标（表7）。每张表均以CSV格式提供，且数据文件与论文中的对应图表相关联。