Handwritten Marathi Digit Recognition remains unsolved due to excessive cursive in Marathi handwriting. The existing classifiers do not give satisfactory performance in real world applications. This paper presents an approach based on Deep Neural Network to improve the performance of Marathi handwritten digit recognition. In this study, Primary database has been used which consists of ten Marathi numerals, A sample was collected from ten different users having different age group and gender for each digit. So hundred samples were collected for the study. The Convolutional Neural Network (CNN) was used to recognize the samples. This network helps to extract feature information, recognition of shapes, scaling and other pattern distortions. Deep Convolutional Network gives better results as compared to traditional classifiers. This approach gives highest average recognition test accuracy of 98 % on our dataset.