DATASET FOR UNIVARIATE TESTS OF OCR SYSTEMS

In this work we present the creation of a repository containing 52,525 synthetic images, generated with features commonly found on digital platforms, and which contains disinformation presented in Portuguese for conducting univariate tests with optical character recognition (OCR) systems. Miscellaneous characteristics in images are: text rotation angle, image dimensions, font color, font size, font style, background color and shadow effect on text. More information can be found at the link: MaVILab-UFV/OCR-eval-for-misinformation-SIBGRAPI-2023 (github.com)