On the Role of Discrete Tokenization in Visual Representation Learning

TL;DR

This paper investigates the impact of discrete tokenization in masked image modeling, providing theoretical insights, introducing a new metric TCAS, and proposing a novel tokenizer and MIM method called ClusterMIM that outperforms existing approaches.

Contribution

It offers a theoretical analysis of discrete tokenization in MIM, introduces the TCAS metric, and presents ClusterMIM, a new method with improved performance on benchmarks.

Findings

Discrete tokenization enhances generalization in MIM.

The TCAS metric effectively evaluates tokenization quality.

ClusterMIM achieves superior results on benchmark datasets.

Abstract

In the realm of self-supervised learning (SSL), masked image modeling (MIM) has gained popularity alongside contrastive learning methods. MIM involves reconstructing masked regions of input images using their unmasked portions. A notable subset of MIM methodologies employs discrete tokens as the reconstruction target, but the theoretical underpinnings of this choice remain underexplored. In this paper, we explore the role of these discrete tokens, aiming to unravel their benefits and limitations. Building upon the connection between MIM and contrastive learning, we provide a comprehensive theoretical understanding on how discrete tokenization affects the model's generalization capabilities. Furthermore, we propose a novel metric named TCAS, which is specifically designed to assess the effectiveness of discrete tokens within the MIM framework. Inspired by this metric, we contribute an innovative tokenizer design and propose a corresponding MIM method named ClusterMIM. It demonstrates superior performance on a variety of benchmark datasets and ViT backbones. Code is available at https://github.com/PKU-ML/ClusterMIM.