This is done with help of a K-Nearest Neighbors classifier, a Fully connected neural network and a Convolutional neural network for Persian Handwritten Digit Recognition. The notebooks are available at
notebooks/knn.ipynb, notebooks/fcnn.ipynb and notebooks/convnn.ipynb respectively.
The name of the dataset used is Hoda and you can know more in this link. Contains 60,000 data for training, 20,000 for testing and more than 22,000 as RemainingSamples. The dataset was in .CDB format so for reading, I used HodaDatasetReader from Amir Saniyan.
A neural network...
