This is an organized handwriting database from IAM Handwriting Database.
In the original database, all images are mixed regardless of the writers. The file "forms_for_parsing.txt" is provided for determine the relationship bewteen single handwriting and its writer.
In this repository, 50 writers with the largest amount of handwriting images are collected and the images are categoried arrcording to its writer. Convenient for train/validate/test spliting and labling.
U. Marti and H. Bunke. The IAM-database: An English Sentence Database for Off-line Handwriting Recognition. Int. Journal on Document Analysis and Recognition, Volume 5, pages 39 - 46, 2002.