VML-MOC: Segmenting a multiply oriented and curved handwritten text lines dataset

TL;DR

This paper introduces the VML-MOC dataset of complex, multi-oriented, and curved handwritten text lines and evaluates a Gaussian-based segmentation method that outperforms single-oriented approaches.

Contribution

The paper provides a new challenging dataset for handwritten text line segmentation and demonstrates the effectiveness of a multi-oriented Gaussian segmentation method.

Findings

Achieved 80.96% mean pixel IoU on the VML-MOC dataset.

Multi-oriented Gaussian segmentation outperforms single-oriented methods.

Dataset includes diverse, curved, and multi-oriented handwritten text lines.

Abstract

This paper publishes a natural and very complicated dataset of handwritten documents with multiply oriented and curved text lines, namely VML-MOC dataset. These text lines were written as remarks on the page margins by different writers over the years. They appear at different locations within the orientations that range between 0 and 180 or as curvilinear forms. We evaluate a multi-oriented Gaussian based method to segment these handwritten text lines that are skewed or curved in any orientation. It achieves a mean pixel Intersection over Union score of 80.96% on the test documents. The results are compared with the results of a single-oriented Gaussian based text line segmentation method.