Papers › Spatial-Channel Token Distillation for Vision MLPs

Spatial-Channel Token Distillation for Vision MLPs

23 Jul 2022International Conference on Machine Learning 2022 7archive 2025-07-28

Yanxi Li, Xinghao Chen, Minjing Dong, Yehui Tang, Yunhe Wang, Chang Xu

Recently, neural architectures with all Multi-layer Perceptrons (MLPs) have attracted great research interest from the computer vision community. However, the inefficient mixing of spatial-channel information causes MLP-like vision models to demand tremendous pre-training on large-scale datasets. This work solves the problem from a novel knowledge distillation perspective. We propose a novel Spatial-channel Token Distillation (STD) method, which improves the information mixing in the two dimensions by introducing distillation tokens to each of them. A mutual information regularization is further introduced to let distillation tokens focus on their specific dimensions and maximize the performance gain. Extensive experiments on ImageNet for several MLP-like architectures demonstrate that the proposed token distillation mechanism can efficiently improve the accuracy. For example, the proposed STD boosts the top-1 accuracy of Mixer-S16 on ImageNet from 73.8% to 75.7% without any costly pre-training on JFT-300M. When applied to stronger architectures, e.g. CycleMLP-B1 and CycleMLP-B2, STD can still harvest about 1.1% and 0.5% accuracy gains, respectively.

Paper PDF Code

Code

devrimcavusoglu/std pytorch report

Repository list and official/mentioned flags are the archive's, frozen 2025-07-28. Reachability, where shown, is from one Syntology probe window (2026-09-16 to 2026-09-18); repositories not probed show nothing. GitHub stars are not tracked.

Code Syntology ran Syntology

Not run by Syntology. Nothing on this page verifies that the listed code works.

Tasks

Image Classification Knowledge Distillation

Results from the paper archive 2025-07-28

Task	Dataset	Model	Metric	Value	Rank at snapshot	Leaderboard	Report
Image Classification	ImageNet	ResMLP-B24 + STD	GFLOPs	24.1	#544 of 1060	Archive leaderboard	report
Image Classification	ImageNet	ResMLP-B24 + STD	Number of params	122.6M	#544 of 1060	Archive leaderboard	report
Image Classification	ImageNet	ResMLP-B24 + STD	Top 1 Accuracy	82.4%	#544 of 1060	Archive leaderboard	report
Image Classification	ImageNet	CycleMLP-B2 + STD	GFLOPs	4.0	#577 of 1060	Archive leaderboard	report
Image Classification	ImageNet	CycleMLP-B2 + STD	Number of params	30.1M	#577 of 1060	Archive leaderboard	report
Image Classification	ImageNet	CycleMLP-B2 + STD	Top 1 Accuracy	82.1%	#577 of 1060	Archive leaderboard	report
Image Classification	ImageNet	Mixer-S16 + STD	GFLOPs	4.3	#940 of 1060	Archive leaderboard	report
Image Classification	ImageNet	Mixer-S16 + STD	Number of params	22.2M	#940 of 1060	Archive leaderboard	report
Image Classification	ImageNet	Mixer-S16 + STD	Top 1 Accuracy	75.7%	#940 of 1060	Archive leaderboard	report

Ranks are positions in the archive's leaderboards as they stood at the 2025-07-28 snapshot. Results published since then are not among these rows, so a rank here is not a current standing.

Methods

Introduced by this paper: STD

Average Pooling Dense Connections Dropout Global Average Pooling Knowledge Distillation Layer Normalization MLP-Mixer Residual Connection STD

Report a problem or propose a change · a person checks every report against the paper or source before anything changes; decisions are listed on /corrections