AI Paper Atlas

Research map / Architectures and efficiency / Attention and transformers

PaPr: Training-Free One-Step Patch Pruning with Lightweight ConvNets for Faster Inference

Tanvir Mahmud, Burhaneddin Yaman, Chun-Hao Liu, Diana Marculescu

ECCV 2024 · Architectures and efficiency › Attention and transformers · approach: Hybrid and interpretable ViT designs · 7 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
replace PaPr's class weights in the FC layer with uniform weights to build the patch significance mapPaPr · ImageNet-1kpatch localizationImproved
use PaPr one-step pruning instead of ToMe at similar FLOPsAugReg ViT-B-16 · ImageNet-1k valtop-1 accuracy: 80.38 -> 82.11Improved
add PaPr pruning before ToMeToMe on ViT-B-16 · ImageNet-1k valtop-1 accuracy / FLOPs: 82.34 acc at 8.21 GFLOPs vs ToMe 80.38 at 8.78Improved
use PaPr instead of ToMe on MAE-pretrained modelsMAE ViT-B-16 · ImageNet-1k valtop-1 accuracy: 78.82 -> 82.4Improved
use PaPr on class-token-free ViTs instead of ToMeViT-Medium GAP-16 256 · ImageNet-1k valtop-1 accuracy: 80.21 -> 81.8Improved
use training-free PaPr instead of trained DynamicCNN/DynamicSwinConvNeXt Base-1k, Swin B-1k · ImageNet-1k valtop-1 accuracy: 83.08 -> 82.75 (ConvNeXt); 83.18 -> 81.7 (Swin)Hurt
use ImageNet-22k weights with PaPrConvNeXt Base · ImageNet-1k valtop-1 accuracy: 82.75 -> 84.27Improved
swap MobileOne-S0 proposal for ResNet-152 (42x more FLOPs) (ablation)PaPr on ViT-B · ImageNet-1k valtop-1 accuracy: 0.3% accuracy differenceNo clear effect
use larger video proposal models (X3d-m, ResNet-101 vs smaller) (ablation)PaPr on ViT-B MAE · Kinetics-400 valtop-1 accuracy: 81.11 -> 81.19 (X3d-s to X3d-m, z=0.5)No clear effect
use PaPr (z=0.3) on ViT-L MAE video modelViT-L MAE · Kinetics-400 valtop-1 accuracy: 85.26 -> 84.53 at 598 -> 160 GFLOPsMixed
replace gradient-based CAM localization with PaPr significance map (ablation)patch pruning of ViT-B · qualitative samplesprediction confidenceImproved
use PaPr for training DeiT-SDeiT-S · ImageNet-1ktrain speed: 1x -> 1.6xImproved

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