AIBangla

Automatic handwritten Bangla character recognition (HBCR) is a challenging problem in computer vision due to numerous variation in writing styles of an individual Bangla character and the presence of similarities with other characters in shapes. Considering the complexity of the problem, we need to develop modern convolutional neural network (CNN) for accurate recognition, but unfortunately at present, very few Bangla handwritten dataset contain a large number of image samples for each character suitable for training deep learning-based methods. In this paper, we present AIBangla, a new benchmark image database of isolated handwritten Bangla characters with detailed usage and baseline. Our dataset contains 80,403 hand-written images on 50 Bangla basic characters and 249,911 hand-written images on 171 Bangla unique pattern shapes compound characters written by more than 2,000 unique writers from various institutes across Bangladesh.

自动手写孟加拉文字符识别（HBCR）是计算机视觉领域的一项颇具挑战性的问题，其原因在于孟加拉文单个字符的书写风格存在众多变异，且与其它字符在形状上存在相似性。鉴于该问题的复杂性，我们需要开发现代卷积神经网络（CNN）以实现准确的识别，然而遗憾的是，目前极少数孟加拉文手写数据集包含适用于训练深度学习方法的大量图像样本。在本研究中，我们提出了AIBangla，一个包含独立手写孟加拉文字符的新基准图像数据库，其中详尽地记录了使用情况和基线。我们的数据集包含由超过2,000位来自孟加拉国各研究所的独特作者创作的80,403张50个孟加拉文基本字符的手写图像以及249,911张171个孟加拉文独特图案形状复合字符的手写图像。