TLDR: Fonts constitute an awesome, free and untapped resource for training text image recognition models. Train on synthetic data (fonts) and transfer to handwritten digits.

Citation:

Vinay Uday Prabhu, Sanghyun Han, Dian Ang Yap, Mihail Douhaniaris, Preethi Seshadri and John Whaley, Fonts-2-Handwriting: A Seed-Augment-Train framework for universal digit classification , Proceedings, ICLR 2019 Workshop on Deep Generative Models for Highly Structured Data, May 2019, New Orleans, USA

BibTex:

@InProceedings{Prabhu2019SAT, author = {Vinay Uday Prabhu, Sanghyun Han, Dian Ang Yap, Mihail Douhaniaris, Preethi Seshadri and John Whaley}, journal = {Proceedings, ICLR 2019 Workshop on Deep Generative Models for Highly Structured Data }, pages = {1--11}, title = { Fonts-2-Handwriting: A Seed-Augment-Train framework for universal digit classification }, year = {2019} }