Momentum Contrast for Unsupervised Visual Representation Learning

We present Momentum Contrast (MoCo) for unsupervised visual representation learning. From a perspective on contrastive learning as dictionary look-up, we build a dynamic dictionary with a queue and a moving-averaged encoder. This enables building a large and consistent dictionary on-the-fly that facilitates contrastive unsupervised learning. MoCo provides competitive results under the common linear protocol on ImageNet classification. More importantly, the representations learned by MoCo transfer well to downstream tasks. MoCo can outperform its supervised pre-training counterpart in 7 detection/segmentation tasks on PASCAL VOC, COCO, and other datasets, sometimes surpassing it by large margins. This suggests that the gap between unsupervised and supervised representation learning has been largely closed in many vision tasks.

PDF Abstract CVPR 2020 PDF CVPR 2020 Abstract

Results from the Paper

Task Dataset Model Metric Name Metric Value Global Rank Uses Extra
Training Data
Result Benchmark
Self-Supervised Image Classification ImageNet MoCo (ResNet-50 4x) Top 1 Accuracy 68.6% # 62
Number of Params 375M # 7
Self-Supervised Image Classification ImageNet MoCo (ResNet-50) Top 1 Accuracy 60.6% # 78
Number of Params 24M # 34
Top 1 Accuracy (kNN, k=20) 47.1% # 16
Self-Supervised Image Classification ImageNet MoCo (ResNet-50 2x) Top 1 Accuracy 65.4% # 68
Number of Params 94M # 19
Contrastive Learning imagenet-1k ResNet50 ImageNet Top-1 Accuracy 60.6 # 11
Self-Supervised Image Classification ImageNet (finetuned) MoCo (Resnet-50) Top 1 Accuracy 77.3% # 34
Top 1 Accuracy 77.0% # 36
Image Classification OmniBenchmark MoCoV2 Average Top-1 Accuracy 34.8 # 15