The Monk Line Segmentation (MLS) Dataset

Overview

The MLS dataset available from this page consists of 31 handwritten page scans. The dataset contains medieval, historical and contemporary manuscripts, and has the purpose of testing line-segmentation algorithms. The collection contains a wide variation of the common problems in handwriting recognition: lines with overlapping ascenders/descenders, slightly rotated scans and curved base lines. 
 

Download

The MLS dataset was collected from the Monk system as of Friday May 17 14:15:04 CEST 2013. It was collected by Lambert Schomaker in May 2013 at the Institution of Artificial Intelligence and Cognitive Engineering (ALICE), University of Gronigen. 

The tar.gz file contains the image dataset for historical manuscripts. For more details please refer to the README file in the tar.gz file. The dataset downloaded for research use only. © 2013 Copyright. 
 

@INPROCEEDINGS{Surinta:2014:ICFHR,
     author = {O. Surinta and M. Holtkamp and M. F. Karaaba and JP. van Oosten and L. R. B. Schomaker and M. A. Wiering},
     title = {A* Path Planning for Line Segmentation of Handwritten Documents},
    booktitle = {Frontiers in Handwriting Recognition (ICFHR), 2014 14th International Conference on},
     year = {2014},
    month = {Sep},
     pages = {175-180},
     numpages = {6},
     isbn = {978-1-4799-4335-7},
    issn = {2167-6445},
    publisher = {IEEE},
     doi = {http://dx.doi.org/10.1109/ICFHR.2014.37},
}