AI Paper Atlas

Research map / Architectures and efficiency / Attention and transformers

DS-Net++: Dynamic Weight Slicing for Efficient Inference in CNNs and Vision Transformers

Changlin Li, Guangrun Wang, Bing Wang, Xiaodan Liang, Zhihui Li, Xiaojun Chang

TPAMI 2022 · Architectures and efficiency › Attention and transformers · approach: Token merging and masked pretraining · 76 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 generalized residual learning instead of naive weight sharing (ablation)dynamic depth supernet · MobileNet V1, 2 pathsaccuracy psi_2: 50.0 -> 53.2Improved
use generalized residual learning instead of naive weight sharing (ablation)dynamic depth supernet · MobileNet V1, 2 pathsaccuracy psi_1: 46.1 -> 48.3Improved
use weight slicing instead of masking (ablation)dynamic ResNet-50 25% channels · latency RTX 2080 Tilatency: 12.4ms -> 7.9 msImproved
use weight slicing instead of indexing (ablation)dynamic ResNet-50 25% channels · latency RTX 2080 Tilatency: 16.6 ms -> 7.9 msImproved
replace LV-ViT stem with depthwise separable stem (ablation)DS-ViT++ patch embedding · ViT stem complexityMAdds: 6.8G -> 5.6GImproved
transfer the trained routing gate to CIFAR-10 (ablation)DS-ResNet · CIFAR-10accuracy: +0.4%Improved
use DS-MBNet-M++ instead of DS-MBNet-M (++ improvements)DS-Net · ImageNet MobileNet 300Mtop-1: 72.8 -> 73.2Improved
use DS-MBNet-M++ over static AutoSlimDS-MBNet · ImageNet MobileNet 300Mtop-1: 71.5 -> 73.2Improved
use DS-ViT-Ti++ over statically scaled DeiT-TiDS-ViT++ · ImageNettop-1: 72.1 -> 78.2Improved

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