ImageNet Large Scale Visual Recognition Challenge
A benchmark for low-shot recognition was proposed by Hariharan & Girshick (2017) and consists of a representation learning phase without access to the low-shot classes and a subsequent low-shot learning phase.
BibTex: