SVC-onGoing:签名验证竞赛
计算机视觉与模式识别
2022-02-23 v2 人机交互
摘要
本文介绍 SVC-onGoing,一项在线签名验证的持续性竞赛,研究者可借助大规模公开数据库(如 DeepSignDB 与 SVC2021_EvalDB)和标准实验协议,在开放公共平台上轻松将其系统与世界最优水平进行基准比对。SVC-onGoing 基于 ICDAR 2021 在线签名验证竞赛(SVC 2021),并已扩展为允许参与者随时加入。SVC-onGoing 的目标是通过大规模公开数据库评估在线签名验证系统在流行场景(办公/移动)与书写输入(触控笔/手指)下的极限。竞赛考虑三项不同任务,模拟真实场景,每项任务同时包含随机伪造与熟练伪造。SVC-onGoing 中获得的结果证明了深度学习方法相较传统方法的高潜力。特别地,最佳签名验证系统取得的等错误率(EER)为 3.33%(任务 1)、7.41%(任务 2)和 6.04%(任务 3)。该领域未来研究应致力于提升签名验证系统在 SVC-onGoing 中具有挑战性的移动场景下的性能,其中签名采集使用了多种移动设备与手指。
引用
@article{arxiv.2108.06090,
title = {SVC-onGoing: Signature Verification Competition},
author = {Ruben Tolosana and Ruben Vera-Rodriguez and Carlos Gonzalez-Garcia and Julian Fierrez and Aythami Morales and Javier Ortega-Garcia and Juan Carlos Ruiz-Garcia and Sergio Romero-Tapiador and Santiago Rengifo and Miguel Caruana and Jiajia Jiang and Songxuan Lai and Lianwen Jin and Yecheng Zhu and Javier Galbally and Moises Diaz and Miguel Angel Ferrer and Marta Gomez-Barrero and Ilya Hodashinsky and Konstantin Sarin and Artem Slezkin and Marina Bardamova and Mikhail Svetlakov and Mohammad Saleem and Cintia Lia Szucs and Bence Kovari and Falk Pulsmeyer and Mohamad Wehbi and Dario Zanca and Sumaiya Ahmad and Sarthak Mishra and Suraiya Jabin},
journal= {arXiv preprint arXiv:2108.06090},
year = {2022}
}
备注
arXiv admin note: substantial text overlap with arXiv:2106.00739