Afro-MNIST: Synthetic generation of MNIST-style datasets for low-resource languages
This work addresses the problem of limited educational resources in underrepresented nations by providing accessible datasets for machine learning education, though it is incremental as it extends an existing dataset format to new domains.
The authors tackled the lack of MNIST-style datasets for low-resource languages by creating Afro-MNIST, a synthetic dataset for four Afro-Asiatic and Niger-Congo orthographies, which serves as a drop-in replacement for MNIST and includes an open-source generation method.
We present Afro-MNIST, a set of synthetic MNIST-style datasets for four orthographies used in Afro-Asiatic and Niger-Congo languages: Ge`ez (Ethiopic), Vai, Osmanya, and N'Ko. These datasets serve as "drop-in" replacements for MNIST. We also describe and open-source a method for synthetic MNIST-style dataset generation from single examples of each digit. These datasets can be found at https://github.com/Daniel-Wu/AfroMNIST. We hope that MNIST-style datasets will be developed for other numeral systems, and that these datasets vitalize machine learning education in underrepresented nations in the research community.