ishtiakm / bornomala

In this project, we have built a database of Bangla Handwritten Letters which contains handwritten images of 84 Bangla letters (10 numerals, 11 vowels, 39 consonants, 24 compound letters). We also investigated some of the existing Bangla character recognition models and found that these models have lower accuracy when the database contains some compound characters. So, we also built a Convolutional Neural Network model to recognize the handwritten Bangla letters and found that our model has higher performance than a lot of existing works.

Bornomala

If you use any code or data from this project, please cite this - DOI: 10.13140/RG.2.2.19658.70082