IndoSign: Indonesian Offline Signature Dataset for Signature Forgery Detection

Description

IndoSign is a balanced offline handwritten signature dataset developed to support signature forgery detection and verification research for Indonesian writers. This dataset contains 2.400 signature images from 60 target writers aged 18 - 60, comprising 1.200 genuine signatures (20 per writer) and 1.200 skilled forgeries produced through a cross-forgery procedure. This dataset is released under a CC BY-NC-ND 4.0 license and is subjected to additional restrictions arising from the informed consent obtained from participating writers. Access is restricted to non-commercial academic research and education, and redistribution of the dataset (in original or modified form) is not permitted without prior written permision from the Provider. Research requesting access must review and sign the accompanying Data Use Agreeent (DUA), including in the files section, and return the signed copy to the corresponding author prior to the receiving access to the dataset.

Steps to reproduce

The IndoSign dataset was constructed by recruiting 60 target writers with diverse background, including students and working professionals aged 18-60 years. Each target writer was then asked to provide 20 genuine handwritten signatures on paper, resulting in 1.200 genuine signature images. To generate forgeries, each participant was randomly assigned to imitate the signatures of two different target writers after observing genuine samples from those writers, with no participant permitted to imitate their own signature. The two forgers assigned to each target writer each produced 10 imitation signatures, resulting in 20 forged signatures per target writer and 1.200 forged images overall.

Institutions

• Binus University

  Jakarta, Jakarta

Categories

Licence