Telegram digits dataset

This dataset is MNIST-like, containing digitized handwritten characters extracted from electoral telegrams during the General Elections of Santa Fe, Argentina, in the year 2021. The dataset offers a valuable resource for researchers and practitioners in the field of character recognition, particularly in the context of electoral data analysis. Each sample in the dataset represents a single digit, ranging from 0 to 9, handwritten by different individuals participating in the electoral process. The dataset aims to facilitate the development and evaluation of machine learning and computer vision algorithms for character recognition tasks.

It contains 170718 images, splitted in train (119502), validation (25608) and test (25608).

This dataset is part of master's thesis in data science which aims to build an Optical Character Recognition (OCR) system using domain adaptation techniques titled "Classification of digits written in the telegrams of legislative elections in Santa Fe using Domain Adaptation techniques".