Paul, Sukriti and Shetty, Nisha P (2018) Handwritten Hindi Numeral Recognition Using Clustering Techniques. International Journal of Engineering and Technology(UAE), 7 (4.41). pp. 145-162. ISSN 2227-524X

PDF 5574.pdf - Published Version Restricted to Registered users only Download (245kB) | Request a copy

Abstract

The problem of automated Hindi numeral recognition is a challenging task owing to the complexity of the script which is characterized by concavities, holes and curvatures. In case of handwritten numerals, the varying writing styles of individuals have to be considered. Our paper focuses at tackling the Hindi numeral recognition problem via various clustering techniques and evaluating them. Subsequently, we work on modifying the framework in Joint Unsupervised Learning (JULE) of Deep Representations and Image Clusters, with different convolutional neural network (CNN) architectures, to obtain normalized mutual information (NMI) results which are better than the state of the art results. Additionally, clustering results obtained on applying different de-noising and contrast adjustment techniques have been presented

Actions (login required)

View Item