Deep learning based optimization model for document layout and text recognition
R. Sundara Rajan, M.S. Geetha Devasena
Abstract
In this study, we use deep learning approaches to offer a novel method for layout anchor box recognition and text analysis in scanned documents. Due to differences in layout, picture quality, and text orientations, scanned documents sometimes provide difficulties. As a result, our goal is to create a reliable deep learning model that can recognize anchor boxes and extract important data from scanned papers. In this study, we introduced the DeepDoc method, a deep learning-based strategy for analyzing document layouts. First, DeepDoc detects semantic structure of document including abstract, title etc. Then, the data is preprocessed and fed into optimal feature selection approach based on Coati’s Optimization Algorithm (COA). The YOLOv3 used to analyze the document completely based on the optimum features learned by COA algorithm. The proposed deep learning model outperforms existing approaches and shows promising solution for document analysis, archiving, and information retrieval.