AI Paper Atlas

Research map / Architectures and efficiency / Attention and transformers

MixMAE: Mixed and Masked Autoencoder for Efficient Pretraining of Hierarchical Vision Transformers

Jihao Liu, Xin Huang, Jinliang Zheng, Yu Liu, Hongsheng Li

CVPR 2023 · Architectures and efficiency › Attention and transformers · approach: Token merging and masked pretraining · 73 citations (OpenAlex)

This paper tried a common change: Train longer

Train longer, from 300 to 600 epochs · ImageNet-1K · top-1 accuracy: 84.6 -> 85.1 Improved

What 387 accepted papers report for this change

See this change in detailMatched by the wording of each paper's own result, read by AI from the full text. Papers report gains more often than failures, so this shows what gets published, not the chance a change works.

Read the paper

AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.

Measured effects reported in the paper

What was changedWhereMeasuredEffect
Use mixed visible tokens and dual reconstruction instead of single-image masked reconstruction (ablation)MixMAE with Swin-B/W14 · ImageNet-1Ktop-1 accuracy: 84.0 -> 84.6Improved
Use mixed visible tokens and dual reconstruction instead of single-image masked reconstruction (ablation)MixMAE with Swin-B/W14 · ADE20KmIoU: 47.3 -> 49.9Improved
Fill masked areas with visible tokens from another image rather than zeros (ablation)MixMAE · ImageNet-1Ktop-1 accuracy: 84.1 -> 84.6Improved
Fill masked areas with visible tokens from another image rather than shuffled tokens (ablation)MixMAE · ImageNet-1Ktop-1 accuracy: 82.6 -> 84.6Improved
Fill masked areas with visible tokens from another image rather than zoomed-and-cropped image tokens (ablation)MixMAE · ImageNet-1Ktop-1 accuracy: 83.5 -> 84.6Improved
Use a shared learnable token rather than zeros for filling (ablation)MixMAE · ADE20KmIoU: 48.0 -> 48.9Improved
Mix 4 images (0.75 masking ratio) rather than 2 images (0.5) (ablation)MixMAE · ImageNet-1Ktop-1 accuracy: 84.6 -> 84.8Improved
Use masked self-attention to separate image groups (ablation)MixMAE · ImageNet-1Ktop-1 accuracy: 84.4 -> 85.1Improved
Train longer, from 300 to 600 epochs (ablation)MixMAE · ImageNet-1Ktop-1 accuracy: 84.6 -> 85.1Improved
Train longer, from 600 to 900 epochs (ablation)MixMAE · ImageNet-1Ktop-1 accuracy: 85.1 -> 85.1No clear effect
Train longer, from 300 to 900 epochs (ablation)MixMAE · ADE20KmIoU: 49.9 -> 51.0Improved
Replace mixed-image filling with [MASK] tokens for a 0.75 masking ratio (ablation)MixMAE · ImageNet-1Ktop-1 accuracy: 84.8 -> 84.4Hurt

Earlier methods this paper compares with

Named baselines from earlier accepted papers in this paper's results tables: better than 1 of 1 on most shared results.

Matched by method name in results tables read by AI. Authors' own tables tend to favour the newer method, and a comparison can be missed when a method is renamed.

Closest papers by meaning

Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.

Open scime draft

Get the next scime Dispatch and a monthly field update

Free. We send a confirmation link first; every email has a one-click unsubscribe. Privacy