Contrastive unsupervised learning has recently shown encouraging progress,
e.g., in Momentum Contrast (MoCo) and SimCLR. In this note, we verify the
effectiveness of two of SimCLR's design improvements by implementing them in
the MoCo framework. With simple modifications to MoCo---namely, using an MLP
projection head and more data augmentation---we establish stronger baselines
that outperform SimCLR and do not require large training batches. We hope this
will make state-of-the-art unsupervised learning research more accessible. Code
will be made public.
Description
Improved Baselines with Momentum Contrastive Learning
%0 Generic
%1 chen2020improved
%A Chen, Xinlei
%A Fan, Haoqi
%A Girshick, Ross
%A He, Kaiming
%D 2020
%K cs.CV
%T Improved Baselines with Momentum Contrastive Learning
%U http://arxiv.org/abs/2003.04297
%X Contrastive unsupervised learning has recently shown encouraging progress,
e.g., in Momentum Contrast (MoCo) and SimCLR. In this note, we verify the
effectiveness of two of SimCLR's design improvements by implementing them in
the MoCo framework. With simple modifications to MoCo---namely, using an MLP
projection head and more data augmentation---we establish stronger baselines
that outperform SimCLR and do not require large training batches. We hope this
will make state-of-the-art unsupervised learning research more accessible. Code
will be made public.
@misc{chen2020improved,
abstract = {Contrastive unsupervised learning has recently shown encouraging progress,
e.g., in Momentum Contrast (MoCo) and SimCLR. In this note, we verify the
effectiveness of two of SimCLR's design improvements by implementing them in
the MoCo framework. With simple modifications to MoCo---namely, using an MLP
projection head and more data augmentation---we establish stronger baselines
that outperform SimCLR and do not require large training batches. We hope this
will make state-of-the-art unsupervised learning research more accessible. Code
will be made public.},
added-at = {2021-09-28T07:04:41.000+0200},
author = {Chen, Xinlei and Fan, Haoqi and Girshick, Ross and He, Kaiming},
biburl = {https://www.bibsonomy.org/bibtex/2504e6be7ba4dd66b9878e7878809b6ef/aerover},
description = {Improved Baselines with Momentum Contrastive Learning},
interhash = {5782384b731338b24e74113adc9bd468},
intrahash = {504e6be7ba4dd66b9878e7878809b6ef},
keywords = {cs.CV},
note = {cite arxiv:2003.04297Comment: Tech report, 2 pages + references},
timestamp = {2021-09-28T07:04:41.000+0200},
title = {Improved Baselines with Momentum Contrastive Learning},
url = {http://arxiv.org/abs/2003.04297},
year = 2020
}