UrbanOccupationsOETR_hdr_Nicaea_6k_dataset

UrbanOccupationsOETR_hdr_Nicaea_6k is the first historical Arabic handwritten digit dataset. It is curated from the first series of Ottoman population registers conducted in the mid-nineteenth century. The dataset was controlled manually and cleaned. It has more than 6000 digits. Five thousand images are divided into the training folder, and the remaining 1000 images are divided into the test folder. Please cite the below paper in your publications if you use the dataset:

Can Y. S., Kabadayı M.E., Curation of Historical Arabic Handwritten Digit Datasets from Ottoman Population Registers: A Deep Transfer Learning Case Study, submitted to IEEE Big Data 2020 Computational Archival Science Workshop.

Download DatasetColored

DatasetColored: This is the original colored version of the dataset.

Download BlackWhite28

DatasetBlackwhite 28×28: This is the black and white 28×28 px version of the dataset.

Download BlackWhite64

DatasetBlackwhite 64×64: This is the black and white 64×64 px version of the dataset.