The Metafont-Glyph dataset (aka "Meggy" dataset) is a collection of glyph images generated from random metafont scripts.

The main focus of this dataset is to help in construction the mappings between glyph images and metafont scripts by data-driven method. The ultimate goal is to develop models that can generate accurate geometric descriptions of glyphs from ancient, non-standardized characters, which can then be used in typesetting systems such as Metafont and Tex.

The target is a different problem from OCR, which is focused on recognizing and transcribing written text from images.