Revisiting Self-Supervised Visual Representation Learning

CVPR 2019 Alexander KolesnikovXiaohua ZhaiLucas Beyer

Unsupervised visual representation learning remains a largely unsolved problem in computer vision research. Among a big body of recently proposed approaches for unsupervised learning of visual representations, a class of self-supervised techniques achieves superior performance on many challenging benchmarks... (read more)

PDF Abstract

Evaluation Results from the Paper


TASK DATASET MODEL METRIC NAME METRIC VALUE GLOBAL RANK COMPARE
Self-Supervised Image Classification ImageNet Revisited Rel.Patch.Loc (ResNet50v1 ×2) Top 1 Accuracy 51.4% # 20
Self-Supervised Image Classification ImageNet Revisited Rel.Patch.Loc (ResNet50v1 ×2) Top 5 Accuracy 74.0% # 14
Self-Supervised Image Classification ImageNet Revisited Rotation (RevNet-50 ×4) Top 1 Accuracy 55.4% # 19
Self-Supervised Image Classification ImageNet Revisited Rotation (RevNet-50 ×4) Top 5 Accuracy 77.9% # 12
Self-Supervised Image Classification ImageNet Revisited Exemplar (ResNet-50v1 ×3) Top 1 Accuracy 46.0% # 22
Self-Supervised Image Classification ImageNet Revisited Exemplar (ResNet-50v1 ×3) Top 5 Accuracy 68.8% # 16
Self-Supervised Image Classification ImageNet Revisited Jigsaw (ResNet50v1 ×2) Top 1 Accuracy 44.6% # 23
Self-Supervised Image Classification ImageNet Revisited Jigsaw (ResNet50v1 ×2) Top 5 Accuracy 68.0% # 17