Abstract

Offline signature verification is one of the most challenging tasks inbiometrics and document forensics. Unlike other verification problems, it needsto model minute but critical details between genuine and forged signatures,because a skilled falsification might often resembles the real signature withsmall deformation. This verification task is even harder in writer independentscenarios which is undeniably fiscal for realistic cases. In this paper, wemodel an offline writer independent signature verification task with aconvolutional Siamese network. Siamese networks are twin networks with sharedweights, which can be trained to learn a feature space where similarobservations are placed in proximity. This is achieved by exposing the networkto a pair of similar and dissimilar observations and minimizing the Euclideandistance between similar pairs while simultaneously maximizing it betweendissimilar pairs. Experiments conducted on cross-domain datasets emphasize thecapability of our network to model forgery in different languages (scripts) andhandwriting styles. Moreover, our designed Siamese network, named SigNet,exceeds the state-of-the-art results on most of the benchmark signaturedatasets, which paves the way for further research in this direction.