AI Paper Atlas

Research map / Label-efficient and robust learning / Self-supervised and contrastive

Bootstrap Masked Visual Modeling via Hard Patch Mining

Haochen Wang, Junsong Fan, Yuxi Wang, Kaiyou Song, Tiancai Wang, Xiangyu Zhang, Zhaoxiang Zhang

TPAMI 2025 · Label-efficient and robust learning › Self-supervised and contrastive · approach: Masked autoencoder pretraining · 10 citations (OpenAlex)

Read the paperCheck your idea against 137,217 papers

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
add auxiliary reconstruction loss predictor (ablation)MAE (RGB) · ImageNet-1K fine-tunetop-1: 82.23 -> 82.49Improved
add learn-to-mask guided by predicted loss (ablation)MAE + L_pred · ImageNet-1K fine-tunetop-1: 82.49 -> 82.95Improved
add HPM to DINO feature distillation (ablation)DINO features · ImageNet-1K fine-tunetop-1: 83.46 -> 84.13Improved
add HPM to CLIP feature distillation (ablation)CLIP features · ImageNet-1K fine-tunetop-1: 83.20 -> 83.58Improved
add HPM to EMA feature target (ablation)EMA features · ImageNet-1K fine-tunetop-1: 82.99 -> 83.47Improved
mask hard patches (argmax) instead of easy patches (argmin) (ablation)HPM · ImageNet-1K fine-tunetop-1: 82.36 -> 82.95Improved
use easy-to-hard schedule instead of hard-to-easy (ablation)HPM · ImageNet-1K fine-tunetop-1: 81.71 -> 82.95Improved
keep randomness: alpha_T 0.5 instead of 1 (ablation)HPM · ImageNet-1K fine-tunetop-1: 82.67 -> 82.95Improved
use relative BCE instead of absolute MSE predictor loss (ablation)HPM · ImageNet-1K fine-tunetop-1: 82.77 -> 82.95Improved
use 8 decoder blocks instead of 1 (ablation)HPM · ImageNet-1K fine-tunetop-1: 82.67 -> 82.95Improved
add HPM to BEiTBEiT · ImageNet-1K fine-tunetop-1: 80.9 -> 81.5Improved
add HPM to MAE (ablation)MAE (RGB) · COCO detectionAP box: 40.45 -> 42.03Improved

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