Improvements in Handwritten and Printed Text Separation in Historical Archival Documents

The presence of handwritten text and annotations combined with typewritten machine-printed in historical archival records make them visually complex, posing challenges for OCR systems accurately transcribing their content. This paper is an extension [1], reporting on improvements the separation from (including typewriters), by use FCN-based models trained datasets created different data synthesis pipelines. Results show a significant increase about 20% intrinsic evaluation artificial test sets, 8% improvement extrinsic subsequent task real documents.