AI Paper Atlas

Research map / Architectures and efficiency / CNN design

MobileViT: Light-weight, General-purpose, and Mobile-friendly Vision Transformer

Sachin Mehta, Mohammad Rastegari

ICLR 2022 · Architectures and efficiency › CNN design · approach: Compact CNN architectures · 151 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 MobileViT-S instead of MobileNetv2 as the DeepLabv3 backboneDeepLabv3 · PASCAL VOC 2012 validationmIOU: 75.7 -> 77.1Improved
Use MobileViT-S instead of MNASNet as the SSDLite backboneSSDLite · MS-COCO validationmAP: 23.0 -> 27.7Improved
Use multi-scale sampler instead of standard samplerMobileNetv2-1.0 · ImageNet-1kTop-1 accuracy: 72.1 -> 73.3Improved
Use multi-scale sampler instead of standard samplerResNet-50 · ImageNet-1kTop-1 accuracy: 77.1 -> 78.6Improved
Add the MobileViT-block skip connection (ablation)MobileViT-S · ImageNet-1kTop-1 accuracy: +0.5%Improved
Set patch sizes to 3,3,3 instead of 2,2,2 (ablation)MobileViT-S · ImageNet-1kTop-1 accuracy: 78.4 -> 78.5Improved
Set patch sizes to 4,4,4 instead of 2,2,2 (ablation)MobileViT-S · ImageNet-1kTop-1 accuracy: 78.4 -> 77.6Hurt
Set patch sizes to 8,4,2 instead of 2,2,2 (ablation)MobileViT-S · ImageNet-1kTop-1 accuracy: 78.4 -> 77.3Hurt
Add label smoothing (ablation)MobileViT-S · ImageNet-1kTop-1 accuracy: 78.0 -> 78.3Improved
Add EMA after label smoothing (ablation)MobileViT-S · ImageNet-1kTop-1 accuracy: 78.3 -> 78.4No clear effect
Use weight decay values from 0.1 to 0.0001 (ablation)MobileViT-S · ImageNet-1kvalidation accuracyNo clear effect

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