Bengali-Handwritten-Digit-Recognition

Bengali Handwritten Digit Recognition using Convolutional Neural Networks (CNNs) Handwritten character recognition is an old problem that has been extensively studied for many different languages around the world. Although very high recognition rate has been achieved repeatedly using a variety of methods on the English character set, there exists limited work on the Bengali character set, and most of these achieved recognition accuracies below 90%. Here, that has employed CNNs to the bengali character set, and that, too, achieved 91.147% accuracy.

Bengali is the native language of Bangladesh, and is spoken by 300 million people worldwide, making it the seventh most spoken language in the world by total number of native speakers. Here, I work with the recognition of the 10 basic digit characters only.

Dataset

The dataset was obtained online from Bangla-Isolated-Lekha. More than 15,000 and around 3,000 images respectively are taken as train-set and test-set.

Click on the image.