Full Page Handwriting Recognition via Image to Sequence Extraction

We present a Neural Network based Handwritten Text Recognition (HTR) model architecture that can be trained to recognize full pages of handwritten or printed text without image segmentation. Being on Image Sequence architecture, it extract in an and then sequence correctly imposing any constraints regarding orientation, layout size non-text. Further, also generate auxiliary markup related formatting, content. use character level vocabulary, thereby enabling language terminology subject. The achieves new state-of-art paragraph recognition the IAM dataset. When evaluated scans real world free form test answers - beset with curved slanted lines, drawings, tables, math, chemistry other symbols performs better than all commercially available HTR cloud APIs. It is deployed production as part commercial web application.