MOMA:Distill from Self-Supervised Teachers

TL;DR

MOMA is a self-supervised distillation framework that combines knowledge from contrastive learning and masked image modeling to produce compact, high-performing models efficiently.

Contribution

It introduces a novel method to distill knowledge from pre-trained MoCo and MAE models into a single student, enhancing performance while reducing training costs.

Findings

MOMA achieves competitive results on various benchmarks.

The method reduces training epochs and computational costs.

It effectively combines two self-supervised paradigms for improved representations.

Abstract

Contrastive Learning and Masked Image Modelling have demonstrated exceptional performance on self-supervised representation learning, where Momentum Contrast (i.e., MoCo) and Masked AutoEncoder (i.e., MAE) are the state-of-the-art, respectively. In this work, we propose MOMA to distill from pre-trained MoCo and MAE in a self-supervised manner to collaborate the knowledge from both paradigms. We introduce three different mechanisms of knowledge transfer in the propsoed MOMA framework. : (1) Distill pre-trained MoCo to MAE. (2) Distill pre-trained MAE to MoCo (3) Distill pre-trained MoCo and MAE to a random initialized student. During the distillation, the teacher and the student are fed with original inputs and masked inputs, respectively. The learning is enabled by aligning the normalized representations from the teacher and the projected representations from the student. This simple design leads to efficient computation with extremely high mask ratio and dramatically reduced training epochs, and does not require extra considerations on the distillation target. The experiments show MOMA delivers compact student models with comparable performance to existing state-of-the-art methods, combining the power of both self-supervised learning paradigms. It presents competitive results against different benchmarks in computer vision. We hope our method provides an insight on transferring and adapting the knowledge from large-scale pre-trained models in a computationally efficient way.