New York University Human & machine learning lab

Omniglot data set

Description: The Omniglot dataset contains over 1600 handwritten characters from 50 different alphabets. It's used for studying one-shot learning and for developing more human-like learning algorithms.
Github link

SCAN DATA SET

Description: SCAN is a set of simple language-driven navigation tasks for studying compositional learning and zero-shot generalization.
Github link