Our new X account is live! Follow @wizwand_team for updates
WorkDL logo mark

A Simple Data Mixing Prior for Improving Self-Supervised Learning

About

Data mixing (e.g., Mixup, Cutmix, ResizeMix) is an essential component for advancing recognition models. In this paper, we focus on studying its effectiveness in the self-supervised setting. By noticing the mixed images that share the same source images are intrinsically related to each other, we hereby propose SDMP, short for $\textbf{S}$imple $\textbf{D}$ata $\textbf{M}$ixing $\textbf{P}$rior, to capture this straightforward yet essential prior, and position such mixed images as additional $\textbf{positive pairs}$ to facilitate self-supervised representation learning. Our experiments verify that the proposed SDMP enables data mixing to help a set of self-supervised learning frameworks (e.g., MoCo) achieve better accuracy and out-of-distribution robustness. More notably, our SDMP is the first method that successfully leverages data mixing to improve (rather than hurt) the performance of Vision Transformers in the self-supervised setting. Code is publicly available at https://github.com/OliverRensu/SDMP

Sucheng Ren, Huiyu Wang, Zhengqi Gao, Shengfeng He, Alan Yuille, Yuyin Zhou, Cihang Xie• 2022

Related benchmarks

TaskDatasetResultRank
Image ClassificationCIFAR-100 (test)--
3518
Image ClassificationCIFAR-10 (test)--
3381
Image ClassificationImageNet-1k (val)
Top-1 Acc80
706
Image ClassificationImageNet 1% labeled--
118
Image ClassificationImageNet (10% labels)
Top-1 Acc68
98
Image ClassificationImageNet-A (val)
Accuracy21.1
55
Linear ClassificationImageNet-1K 1.0 (val)
Top-1 Accuracy73.8
48
Linear ClassificationImageNet-1k (val)
Top-1 Accuracy73.5
37
Image ClassificationImageNet-100 small-scale (test)
Top-1 Acc83.2
5
Image ClassificationImageNet-R original (val)
Top-1 Acc45.3
4
Showing 10 of 11 rows

Other info

Code

Follow for update