CVJun 15, 2022

A Simple Data Mixing Prior for Improving Self-Supervised Learning

arXiv:2206.07692v146 citationsh-index: 134Has Code
Originality Highly original
AI Analysis

This addresses the challenge of effectively applying data mixing in self-supervised learning for computer vision, offering a novel approach to boost model performance.

The paper tackled the problem of improving self-supervised learning by leveraging data mixing, proposing SDMP to use mixed images as positive pairs, which enhanced accuracy and out-of-distribution robustness for frameworks like MoCo and enabled Vision Transformers to benefit from data mixing for the first time.

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

Code Implementations1 repo
Foundations

The foundational work for this paper's niche, ranked by how specifically the neighbourhood builds on it — not by global fame.

Your Notes