Record link: 
http://hdl.handle.net/20.500.12708/139868
-
Title: 
Writer Identification and Writer Retrieval Using Vision Transformer for Forensic Documents
en
Citation: 
Koepf, M., Kleber, F., & Sablatnig, R. (2022). Writer Identification and Writer Retrieval Using Vision Transformer for Forensic Documents. In Document Analysis Systems: 15th IAPR International Workshop, DAS 2022, La Rochelle, France, May 22–25, 2022, Proceedings (pp. 352–366). https://doi.org/10.1007/978-3-031-06555-2_24
-
Publisher DOI: 
10.1007/978-3-031-06555-2_24
-
Publication Type: 
Inproceedings - Full-Paper Contribution
en
Language: 
English
-
Authors: 
Koepf, Michael 
Kleber, Florian 
Sablatnig, Robert 
-
Organisational Unit: 
E193-01 - Forschungsbereich Computer Vision 
-
Published in: 
Document Analysis Systems: 15th IAPR International Workshop, DAS 2022, La Rochelle, France, May 22–25, 2022, Proceedings 
-
ISBN: 
978-3-031-06555-2
-
Volume: 
13237
-
Date (published): 
2022
-
Event name: 
15th IAPR International Workshop (DAS 2022) 
en
Event date: 
22-May-2022 - 25-May-2022
-
Event place: 
La Rochelle, France
-
Number of Pages: 
15
-
Peer reviewed: 
Yes
-
Keywords: 
Forensics; Vision Transformer; Writer identification; Writer retrieval
en
Abstract: 
Writer identification and writer retrieval deal with the analysis of handwritten documents regarding the authorship and are used, for example, in forensic investigations. In this paper, we present a writer identification and writer retrieval method based on Vision Transformers. This is in contrast to the current state of the art, which mainly uses traditional Convolutional-Neural-Network-approaches. The evaluation of our self-attention-based and convolution-free method is done on two public datasets (CVL Database and dataset of the ICDAR 2013 Competition on Writer Identification) as well as a forensic dataset (WRITE dataset). The proposed system achieves a top-1 accuracy up to 99% (CVL) and 97% (ICDAR 2013). In addition, the impact of the used script (Latin and Greek) and the used writing style (cursive handwriting and block letters) on the recognition rate are analyzed and presented.
en
Project title: 
IT unterstützte Suche und Vergleich von Handschriften: 879687 (FFG - Österr. Forschungsförderungs- gesellschaft mbH) 
-
Research Areas: 
Visual Computing and Human-Centered Technology: 100%
-
Science Branch: 
1020 - Informatik: 90%
1010 - Mathematik: 10%
-
Appears in Collections:
Conference Paper

Show full item record

Google ScholarTM

Check