EMPIRICAL EVALUATION OF STATE-OF-THE-ART OBJECT DETECTION METHODS FOR DOCUMENT IMAGE UNDERSTANDING

Abstract

The majority of online documents such as research papers, articles, and magazines is publicly available in the image form due to the copyright issue. Document image understanding is the task of deriving a high level presentation of the contents of a document image, which involves several phases, mainly including page segmentation (or block segmentation), blocks classification (or blocks labeling) and several operations for processing text, tables, graphics, figures, formulas, etc. Our objective focuses on the first two phases of document image understanding, namely, locating the logical objects in document pages. This process is valuable for a variety of document image analysis applications. To this end, we evaluate different state-of-the-art object detection methods based on computer vision for the task. Through our extensive experiments, we report findings/comments from the off-the-shelf object detectors and streamline several potential directions for the future work.