面向离线手写签名验证的基于协组损失的多尺度特征学习
Multiscale Feature Learning Using Co-Tuplet Loss for Offline Handwritten Signature Verification
- Department of Information Management, National Taiwan University（信息管理系，国立台湾大学）
- Center for Research in Econometric Theory and Applications, National Taiwan University（计量经济理论与应用研究中心，国立台湾大学）
机构由 AI 辅助整理，请以论文原文为准。
AI总结：
针对离线手写签名验证面临的类间相似、类内差异和样本有限问题，提出带协组损失的多尺度特征学习网络MS-SigNet，同时发布中文签名数据集HanSig，在多语言基准上取得优于现有方法的性能。
AI中文摘要：
手写签名验证对法律和金融机构至关重要，当前面临书写者间相似度高、书写者内差异大以及签名样本有限等挑战。为解决这些问题，我们提出了搭载协组（co-tuplet）损失的多尺度签名特征学习网络（MultiScale Signature feature learning Network, MS-SigNet），这是一种专为离线手写签名验证设计的新型度量学习损失。MS-SigNet从多个空间尺度学习签名的全局和区域特征，提升了特征的区分度。该方法通过捕捉整体笔画和细微的局部差异，能够有效区分真实签名与熟练伪造签名。协组损失聚焦于多个正样本和负样本，通过解决书写者间相似度和书写者内差异问题并突出信息丰富的样本，克服了传统度量学习损失的局限性。代码可在https://github.com/ashleyfhh/MS-SigNet获取。我们还提出了HanSig，一个大规模中文签名数据集，用于支持中文签名鲁棒系统的开发，该数据集可在https://github.com/hsinmin/HanSig访问。在四个不同语言的基准数据集上的实验结果表明，与当前最优方法相比，我们的方法展现出了优异的性能。
英文摘要：
Handwritten signature verification, crucial for legal and financial institutions, faces challenges including inter-writer similarity, intra-writer variations, and limited signature samples. To address these, we introduce the MultiScale Signature feature learning Network (MS-SigNet) with the co-tuplet loss, a novel metric learning loss designed for offline handwritten signature verification. MS-SigNet learns both global and regional signature features from multiple spatial scales, enhancing feature discrimination. This approach effectively distinguishes genuine signatures from skilled forgeries by capturing overall strokes and detailed local differences. The co-tuplet loss, focusing on multiple positive and negative examples, overcomes the limitations of typical metric learning losses by addressing inter-writer similarity and intra-writer variations and emphasizing informative examples. The code is available at https://github.com/ashleyfhh/MS-SigNet. We also present HanSig, a large-scale Chinese signature dataset to support robust system development for this language. The dataset is accessible at https://github.com/hsinmin/HanSig. Experimental results on four benchmark datasets in different languages demonstrate the promising performance of our method in comparison to state-of-the-art approaches.
