The Blog



Share

HEXA: Self-supervised pretraining with hard examples improves visual representations