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

We introduce Bootstrap Your Own Latent (BYOL), a new approach to self-supervised image representation learning. BYOL relies on two neural networks, referred to as online and target networks, that interact and learn from each other... (read more)

PDF Abstract
TASK DATASET MODEL METRIC NAME METRIC VALUE GLOBAL RANK RESULT BENCHMARK
Self-Supervised Image Classification ImageNet BYOL (ResNet-50) Top 1 Accuracy 74.3% # 13
Top 5 Accuracy 91.6% # 9
Self-Supervised Image Classification ImageNet BYOL (ResNet-50 x2) Top 1 Accuracy 77.4% # 6
Top 5 Accuracy 93.6% # 4
Number of Params 94M # 8
Self-Supervised Image Classification ImageNet BYOL (ResNet-50 x4) Top 1 Accuracy 78.6% # 3
Top 5 Accuracy 94.2% # 3
Number of Params 375M # 3
Self-Supervised Image Classification ImageNet BYOL (ResNet-200 x2) Top 1 Accuracy 79.6% # 2
Top 5 Accuracy 94.8% # 2
Number of Params 250M # 5
Semi-Supervised Image Classification ImageNet - 1% labeled data BYOL (ResNet-50 x2) Top 5 Accuracy 84.1% # 11
Top 1 Accuracy 62.2% # 11
Semi-Supervised Image Classification ImageNet - 1% labeled data BYOL (ResNet-50 x4) Top 5 Accuracy 87.9% # 6
Top 1 Accuracy 69.1% # 6
Semi-Supervised Image Classification ImageNet - 1% labeled data BYOL (ResNet-200 x2) Top 5 Accuracy 89.5% # 5
Top 1 Accuracy 71.2% # 5
Semi-Supervised Image Classification ImageNet - 1% labeled data BYOL (ResNet-50) Top 5 Accuracy 78.4% # 15
Top 1 Accuracy 53.2% # 15

Methods used in the Paper