Ubiquitous Document Capturing with Deep-Learning

Abstract

Digital and paper documents co-exist in our daily life. Seamless integration of information from both sources is crucial for e cient knowledge management. The project aims at developing an algorithm which can handle the detection of document so that it can be captured easily to convert it into a digital form for automatic integration of relevant information in electronic work-flows. An approach based on machine learning is proposed unlike the most commonly geometric based methods. Convolutional neural networks (CNN) are used to train a binary classification model and the main algorithm is working by combining line segment detection with CNN model to extract the boundary of the document. CNN model detects the absence and presence of document in the given input and thus that model is then being applied onto the parts of the image being divided along the detected lines. The usage of deep-learning technique provides a solution which is more generalized and flexible than other available solutions.