AI Paper Atlas

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

SdAE: Self-Distillated Masked Autoencoder

Yabo Chen, Yuchen Liu, Dongsheng Jiang, Xiaopeng Zhang, Wenrui Dai, Hongkai Xiong, Qi Tian

ECCV 2022 · Label-efficient and robust learning › Self-supervised and contrastive · approach: Masked autoencoder pretraining · 70 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 teacher feature normalization (ablation)SdAE with full-image teacher · ImageNet-1k fine-tuning (100 ep pretrain)top-1 accuracy: 83.7 -> 84.0Improved
feed only masked tokens to the teacher instead of the full image (ablation)SdAE · ImageNet-1k fine-tuning (100 ep pretrain)top-1 accuracy: 84.0 -> 83.5Hurt
add multi-fold masking over masked-only teacher input (ablation)SdAE · ImageNet-1k fine-tuning (100 ep pretrain)top-1 accuracy: 83.5 -> 84.1Improved
replace full-image teacher input with multi-fold masking (ablation)SdAE · ImageNet-1k fine-tuning (100 ep pretrain)top-1 accuracy: 84.0 -> 84.1Improved
replace full-image teacher input with multi-fold masking (ablation)SdAE · MAE-based training costtraining time per epoch (s): 826s -> 706sImproved
feed the whole image instead of 36 patches to the teacher (ablation)SdAE teacher crop · ImageNet-1k fine-tuning (100 ep pretrain)top-1 accuracy: 82.81 -> 83.04Improved
replace MAE pixel targets with self-distilled latent targetsMAE · ImageNet-1k fine-tuning, ViT-B 300 eptop-1 accuracy: 82.9 -> 84.1Improved
replace MAE pixel targets with self-distilled latent targetsMAE · ADE20K, 300 epmIoU: 45.8 -> 48.6Improved
replace MAE pixel targets with self-distilled latent targetsMAE · COCO detection, 300 epbox AP: 45.4 -> 48.9Improved
pretrain longer, 300 to 800 epochsSdAE · ImageNet-1k fine-tuning, ViT-Btop-1 accuracy: 84.1 -> 84.0Hurt
use a 2-layer decoder instead of 8 layers (ablation)SdAE · ImageNet-1k fine-tuning (100 ep pretrain)top-1 accuracy: 82.76No clear effect
update teacher with EMA every iteration instead of every epoch (ablation)SdAE · ImageNet-1k fine-tuningtop-1 accuracyHurt

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