Single-Writer Strikethrough Dataset

This dataset contains registered pairs of clean and struck-through handwritten words which can be used for the task of (unpaired) strikethrough removal. The text, a passage from Bram Stoker's Dracula, was written by a single writer, using a blue ballpoint pen on regular white paper. The 756 word images have been systematically struck through, using one of the following stroke types: single horizontal line, double horizontal lines, diagonal, cross, wave, zigzag, scratch.

The dataset has been split into three subsets, train, validation and test, each balanced with regard to the number of samples per stroke type. Each split contains csv-files that indicate the stroke type that has been applied to a particular image.