Bangla Handwritten Digit Recognition

Due to the wide range of shapes, sizes, and writing styles, handwritten digit recognition has always been difficult. For all of the recognition algorithms built in this thesis, the rewarded method is used. NumtaDB is the largest dataset. It is a collection for Bengali handwritten digits. This is a massive dataset with over 85,000 images in it. This dataset, however, is incredibly difficult to work with due to its complexity. The aim of this paper is to preprocess images that can be used to train deep learning models with high accuracy. The reason for this is that, unlike the MNIST for English digits, there is no preprocessed data to deal with when it comes to Bengali digit recognition. This unbiased dataset, NumtaDB, is used in this paper for Bangla digit recognition. In this paper, various preprocessing techniques are developed for image processing, with a deep convolutional neural network (CNN) functioning as the classification algorithm. On the NumtaDB image dataset, the performance is systematically evaluated of this process. Finally, 93% accuracy is obtained for NumtaDB dataset, and 92% accuracy rate is obtained for our own dataset using same of the proposed method in experiments. In this paper, the accuracy for every training dataset in NumtaDB is separately calculated which has the maximum accuracy 98%.