Generative Pipeline for Data Augmentation of Unconstrained Document Images with Structural and Textural Degradation (Student Abstract)

Computer vision applications for document image understanding (DIU) such as optical character recognition, word spotting, enhancement etc. suffer from structural deformations like strike-outs and unconstrained strokes, to name a few. They also texture degradation due blurring, aging, or blotting-spots The DIU with deep networks are limited constrained environment lack diverse data text-level pixel-level annotation simultaneously. In this work, we propose generative framework produce realistic synthetic handwritten images simultaneous of text corresponding spatial foreground information. proposed approach generates backgrounds artificial texts which supplements data-augmentation in multiple systems. is an early work facilitate system-evaluation both quality recognition performance at go.