Data sets for training, validation and testing are now available (Please click here)
Identification of writers from handwritten samples is a well-known behavioral biometric modality that finds applications in a variety of problem areas including forensic document analysis, authentication of documents and verification of the genuineness of historical manuscripts etc. Writer identification has been an active area of research during the recent years and a number of identification system realizing promising results have been reported in the literature. Most of the research on writer identification and related tasks, however, has been carried out on handwritten texts in a single script. It is a well-known fact that in most cultures, at least two languages are spoken and written by a significant proportion of the population. Having writing samples of same individuals in multiple scripts allows studying the interesting problem of multi-script writer identification where writing patterns that are common across different scripts may be exploited to identify the writer.
In order to objectively compare the performance of recent advances in multi-script writer identification and to investigate the performance of traditional script-dependent writer identification systems in a multi-script environment, we propose to organize this third edition of Multi-script Writer Identification Competition using LAMIS-MSHD and CERUG Data sets, in conjunction with the 14th International Conference on Document Analysis and Recognition, ICDAR 2017.
The underlying objective of organizing this competition is to study the interesting writer identification scenario when training and test samples of an individual are in two different scripts. The competition will also provide researchers with the opportunity to compare their algorithms under same experimental settings and evaluation protocols and, study the performance evolution of traditional identification systems in a more challenging experimental setup. A report on the competition will be published in the proceedings of ICDAR 2017.The report will comprise description of the participating methods, the evaluation protocol and the final rankings of the participating algorithms. The results of the competition will also be presented in a dedicated session at ICDAR 2017.