Page Layout Analysis System for Unconstrained Historic Documents

Extraction of text regions and individual lines from historic documents is necessary for automatic transcription. We propose extending a CNN-based baseline detection system by adding line height block boundary predictions to the model output, allowing extract more comprehensive layout information. also show that pixel-wise orientation prediction can be used processing with multiple orientations. demonstrate proposed method performs well on cBAD dataset. Additionally, we benchmark newly introduced PERO dataset which make public.