AI Paper Atlas

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

ResMLP: Feedforward Networks for Image Classification With Data-Efficient Training

Hugo Touvron, Piotr Bojanowski, Mathilde Caron, Matthieu Cord, Alaaeldin El-Nouby, Édouard Grave, Gautier Izacard, Armand Joulin, Gabriel Synnaeve, Jakob Jozef Verbeek, Hervé Jeǵou

TPAMI 2022 · Label-efficient and robust learning › Self-supervised and contrastive · approach: Invariance-based self-supervised learning · 952 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
use LayerNorm instead of the Aff affine transform (ablation)ResMLP-S12 · ImageNet-1k valtop-1 accuracy: 76.6 -> 77.7Improved
use Class-MLP pooling instead of average pooling (ablation)ResMLP-S12 · ImageNet-1k valtop-1 accuracy: 76.6 -> 77.5Improved
add the cross-patch linear layer (ablation)ResMLP-S12 · ImageNet-1k valtop-1 accuracy: 56.5 -> 76.6Improved
use 8x8 patches instead of 16x16 (ablation)ResMLP-S12 · ImageNet-1k valtop-1 accuracy: 76.6 -> 79.1Improved
add distillation from a convnet (ablation)ResMLP-S12 · ImageNet-1k valtop-1 accuracy: 76.6 -> 77.8Improved
add distillation from a convnet (ablation)ResMLP-B24/8 · ImageNet-1k valtop-1 accuracy: 81.0 -> 83.6Improved
use the DeiT training recipe instead of old ResNet-style 90-epoch training (ablation)ResMLP-S12 · ImageNet-1k valtop-1 accuracy: 69.2 -> 76.6Improved
pre-train on ImageNet-21k (ablation)ResMLP-B24/8 · ImageNet-1k valtop-1 accuracy: 81.0 -> 84.4Improved
add RandAugment (ablation)ResMLP-S12 · ImageNet-1k valtop-1 accuracy: 73.2 -> 76.6Improved
add DINO self-supervised pre-training before supervised fine-tuning (ablation)ResMLP-S24 · ImageNet-1k valtop-1 accuracy: 79.4 -> 79.9Improved
train for 800 epochs instead of 300 (ablation)ResMLP-S36 · ImageNet-1k valtop-1 accuracy: 79.3 -> 80.4Improved
use SiLU instead of GELU (ablation)ResMLP (CaiT-derived MLP baseline) · ImageNet-1k valtop-1 accuracy: 79.2 -> 78.7Hurt

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