PR-MIM: Delving Deeper into Partial Reconstruction in Masked Image Modeling

TL;DR

This paper introduces PR-MIM, a novel partial reconstruction method for masked image modeling that maintains high performance while significantly reducing computational costs by reconstructing thrown tokens in a lightweight manner.

Contribution

It proposes a progressive and furthest sampling reconstruction strategy that improves partial reconstruction efficiency without sacrificing representation quality.

Findings

Achieves lossless performance with 50% patches thrown

Saves 28% FLOPs and 36% memory compared to standard MAE

Validates effectiveness across various frameworks

Abstract

Masked image modeling has achieved great success in learning representations but is limited by the huge computational costs. One cost-saving strategy makes the decoder reconstruct only a subset of masked tokens and throw the others, and we refer to this method as partial reconstruction. However, it also degrades the representation quality. Previous methods mitigate this issue by throwing tokens with minimal information using temporal redundancy inaccessible for static images or attention maps that incur extra costs and complexity. To address these limitations, we propose a progressive reconstruction strategy and a furthest sampling strategy to reconstruct those thrown tokens in an extremely lightweight way instead of completely abandoning them. This approach involves all masked tokens in supervision to ensure adequate pre-training, while maintaining the cost-reduction benefits of partial reconstruction. We validate the effectiveness of the proposed method across various existing frameworks. For example, when throwing 50% patches, we can achieve lossless performance of the ViT-B/16 while saving 28% FLOPs and 36% memory usage compared to standard MAE. Our source code will be made publicly available