Self-Supervised Learning of Pretext-Invariant Representations

4 Dec 2019Ishan MisraLaurens van der Maaten

The goal of self-supervised learning from images is to construct image representations that are semantically meaningful via pretext tasks that do not require semantic annotations for a large training set of images. Many pretext tasks lead to representations that are covariant with image transformations... (read more)

PDF Abstract
TASK DATASET MODEL METRIC NAME METRIC VALUE GLOBAL RANK COMPARE
Self-Supervised Image Classification ImageNet PIRL Top 1 Accuracy 63.6% # 13
Self-Supervised Image Classification ImageNet PIRL Number of Params 24M # 1
Semi-Supervised Image Classification ImageNet - 10% labeled data PIRL (ResNet-50) Top 5 Accuracy 83.8% # 9