Table 1 Splitting the Arabic characters datasets (80%, 10%, and 10%) for training, validation, and testing sets.

From: Integrating CNN and transformer architectures for superior Arabic printed and handwriting characters classification