Bootstrap your own latent: A new approach to self-supervised Learning

Open paper
Benchmark Model Rank Results
image-classification-on-places205BYOL#12Top 1 Accuracy: 54.0
person-re-identification-on-sysu-30kBYOL (self-supervised)#5Rank-1: 12.7
self-supervised-image-classification-on-imagenetBYOL (ResNet-200 x2)#30Top 1 Accuracy: 79.6%Top 5 Accuracy: 94.8%Number of Params: 250M
self-supervised-image-classification-on-imagenetBYOL (ResNet-50 x4)#40Top 1 Accuracy: 78.6%Top 5 Accuracy: 94.2%Number of Params: 375M
self-supervised-image-classification-on-imagenetBYOL (ResNet-50 x2)#49Top 1 Accuracy: 77.4%Top 5 Accuracy: 93.6%Number of Params: 94M
self-supervised-image-classification-on-imagenetBYOL (ResNet-50)#78Top 1 Accuracy: 74.3%Top 5 Accuracy: 91.6%Number of Params: 24M
self-supervised-person-re-identification-on-sysu-30kBYOL#2Rank-1: 12.7
semi-supervised-image-classification-on-imagenet-1-labeled-dataBYOL (ResNet-200 x2)#14Top 1 Accuracy: 71.2%Top 5 Accuracy: 89.5%
semi-supervised-image-classification-on-imagenet-1-labeled-dataBYOL (ResNet-50 x4)#18Top 1 Accuracy: 69.1%Top 5 Accuracy: 87.9%
semi-supervised-image-classification-on-imagenet-1-labeled-dataBYOL (ResNet-50 x2)#31Top 1 Accuracy: 62.2%Top 5 Accuracy: 84.1%
semi-supervised-image-classification-on-imagenet-1-labeled-dataBYOL (ResNet-50)#44Top 1 Accuracy: 53.2%Top 5 Accuracy: 78.4%

Rank counts only results with a code link.