In proposed work classification of Malayalam handwritten characters using 80 class labels with 1000
instances for each class. Realization of recognition accuracies in handwritten text is an challenging and
never exhausting research problem. The factor‟s which pose challenges in handwritten character
recognition includes high degree of variability in writing especially in Malayalam handwritten script, type
of script and document type are complex and curved nature. For classification a modified CNN architecture
is proposed for which an accuracy of 99.55% is achieved.