{"about":{"site":"https://codewithpapers.app","non_affiliation":"Code with Papers and Syntology are not affiliated with, endorsed by, or sponsored by Papers with Code, Meta, or the pwc-archive mirror.","licence":"CC BY-SA 4.0","licence_url":"https://creativecommons.org/licenses/by-sa/4.0/legalcode","attribution":"https://codewithpapers.app/attribution","modified":"archive material modified by Syntology; see the attribution page"},"url":"/paper/pdnet-semantic-segmentation-integrated-with-a","title":"PDNet: Semantic Segmentation integrated with a Primal-Dual Network for Document binarization","arxiv_id":"1801.08694","date":"2018-01-26","proceeding":null,"authors":["Kalyan Ram Ayyalasomayajula","Filip Malmberg","Anders Brun"],"abstract":"Binarization of digital documents is the task of classifying each pixel in an\nimage of the document as belonging to the background (parchment/paper) or\nforeground (text/ink). Historical documents are often subjected to\ndegradations, that make the task challenging. In the current work a deep neural\nnetwork architecture is proposed that combines a fully convolutional network\nwith an unrolled primal-dual network that can be trained end-to-end to achieve\nstate of the art binarization on four out of seven datasets. Document\nbinarization is formulated as an energy minimization problem. A fully\nconvolutional neural network is trained for semantic segmentation of pixels\nthat provides labeling cost associated with each pixel. This cost estimate is\nrefined along the edges to compensate for any over or under estimation of the\nforeground class using a primal-dual approach. We provide necessary overview on\nproximal operator that facilitates theoretical underpinning required to train a\nprimal-dual network using a gradient descent algorithm. Numerical instabilities\nencountered due to the recurrent nature of primal-dual approach are handled. We\nprovide experimental results on document binarization competition dataset along\nwith network changes and hyperparameter tuning required for stability and\nperformance of the network. The network when pre-trained on synthetic dataset\nperforms better as per the competition metrics.","url_abs":"http://arxiv.org/abs/1801.08694v3","url_pdf":"http://arxiv.org/pdf/1801.08694v3.pdf","source":{"archive":"pwc-archive (Hugging Face), CC BY-SA 4.0","snapshot":"2025-07-28","licence_url":"https://creativecommons.org/licenses/by-sa/4.0/legalcode","row_kind":"abstracts"},"code_links":[{"paper_slug":"pdnet-semantic-segmentation-integrated-with-a","repo_url":"https://github.com/krayyalasomayajula/pdNet","is_official":1,"mentioned_in_paper":1,"mentioned_in_github":0,"framework":"none","reach":null}],"tasks":[{"task_slug":"binarization","task_name":"Binarization"},{"task_slug":"semantic-segmentation","task_name":"Semantic Segmentation"}],"methods":[],"datasets_introduced":[],"methods_introduced":[],"results":[],"syntology":{"atlas_url":null,"mcp":null,"developers":"https://syntology.ai/developers"},"arxiv_metadata":null,"syntology_extracted_results":null}