Brazilian Forensic Letter Database

 The BFL database composed of 315 writers, three samples per writer, summing up 945 images. The motivation for using such database is the growing interest of the Brazilian forensic experts as well as the Brazilian Federal Police in writer verification/identification .
The samples were provided by undergraduate students inthree different sessions during one month. The texts werecollected on an A4 white sheet of paper with no pen-drawbaseline and then scanned in gray level with 300 dpi (3760 X 2448). Each writer was allowed to use his/her own pen, which means that several different pens were used. The text is concise (131 words) and complete in the sense that it contains all characters (letters and numerals) and certain character combinations of interest. This makes it suitable for text-dependent writer identification as well. The image below shows the letter contents and a sample of the database.

How to obtain access to the images

The BFL Database may be used for non-commercial research provided you acknowledge the source of the image by citing the following paper in publications about your research:

  • C. Freitas, L. S. Oliveira, R. Sabourin, F. Bortolozzi Brazilian Forensic Letter Database, 11th International Workshop on Frontiers on Handwriting Recognition (IWFHR-11), Montreal, QC, Canada, August 19-21 2008. (pdf).

Before you can download the database we ask you to register by sending us an email (vri AT inf.ufpr.br) so we are aware of who is using our data. Please include in your email your Last name, First name, Organization and email. Once you have registered you will receive a link to download the database.

Related papers

Our last results on this database can be found in these references

  • Hanusiak, R. K., Justino, E., Oliviera, L. S., Sabourin, R. Writer Verification Using Texture-based Features, International Journal of Document Analysis and Recognition, 15(3):213-226, 2012. (pdf)
  • Bertolini, D.,Oliveira, L. S., Justino, E., Sabourin, R., Texture-based Descriptors for Writer Identification and Verification, Expert Systems With Applications, 40(6):2069-2080, 2013. (pdf).

This database is licensed under a Creative Commons Attribution 4.0 International License.