Abstract

The dataset contains binary handwriting masks, which are sampled from scanned pages. Based on the overlay size, the training and test datasets are divided into five size ranges.

Links and resources

Tags