randall-lab/arabic-digits

阿拉伯数字手写识别数据集，包含700位参与者书写的70000个阿拉伯数字手写图像。数据集分为训练集60000张图像和测试集10000张图像，涵盖了10个阿拉伯数字（标记为0-9）。每个数字由每位书写者书写了10次，确保了手写风格的多样性。图像为灰度格式，分辨率为28x28像素，来源于MADBase数据库。该数据集遵循开放数据库许可证（ODbL）。

This dataset contains 70,000 Arabic handwritten digits written by 700 participants, intended for Arabic digit recognition tasks using machine learning. It is split into a training set of 60,000 images and a test set of 10,000 images, covering 10 Arabic digits (labeled 0-9), with each digit written ten times by each writer. The images are in grayscale, 28x28 pixels, and collected from different institutions to ensure diversity in handwriting styles. The dataset is derived from the MADBase database and is licensed under the Open Database License (ODbL).