BYOL

Reported on 5 benchmarks across 4 tasks · 2 papers · 1 SOTA

Note: results are matched by exact model name. Different papers may use the same name for different model variants.

Computer Vision3 results

Image ClassificationonPlaces205
Top 1 Accuracy· 2020-06-13
54
best: 69.3 (MixMIM-L)
SOTA
Bootstrap your own latent: A new approach to self-supervised Learning arXiv:2006.07733
Image ClassificationonObjectNet
Top-1 Accuracy· uses extra data· 2022-01-13
23
best: 82.7 (CoCa)
Pushing the limits of self-supervised ResNets: Can we outperform supervised learning without labels on ImageNet?arXiv:2201.05119
Person Re-IdentificationonSYSU-30k
Rank-1· 2020-06-13
12.7
best: 36 (IICS (generalization))
Bootstrap your own latent: A new approach to self-supervised Learning arXiv:2006.07733

Medical1 result

Semantic SegmentationonCityscapes val
mIoU· uses extra data· 2022-01-13
74.6
best: 90.3 (EfficientPS (Cityscapes-fine))
Pushing the limits of self-supervised ResNets: Can we outperform supervised learning without labels on ImageNet?arXiv:2201.05119

Audio1 result

10-shot image generationonCityscapes val
mIoU· uses extra data· 2022-01-13
74.6
best: 90.3 (EfficientPS (Cityscapes-fine))
Pushing the limits of self-supervised ResNets: Can we outperform supervised learning without labels on ImageNet?arXiv:2201.05119