AI Paper Atlas

Research map / Architectures and efficiency / CNN design

Refining Activation Downsampling With SoftPool

Alexandros Stergiou, Ronald Poppe, Grigorios Kalliatakis

ICCV 2021 · Architectures and efficiency › CNN design · approach: Filter design and trained-filter analysis · 317 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 max pooling with SoftPool, trained from scratchResNet18 · ImageNet1K val (from scratch)top-1 accuracy: 69.76 -> 71.27Improved
replace max pooling with SoftPool, trained from scratchResNeXt101 32x8d · ImageNet1K val (from scratch)top-1 accuracy: 79.31 -> 80.12Improved
swap in SoftPool on a pre-trained network without retrainingResNet18 · ImageNet1K val (pre-trained, pooling swapped)top-1 accuracy: 69.76 -> 70.56Improved
train from scratch with SoftPool instead of only swapping it into a pre-trained networkResNet18 · ImageNet1K valtop-1 accuracy: 70.56 -> 71.27Improved
replace pooling with SoftPool rather than LIPResNet18 · ImageNet1K val (from scratch, pooling comparison)top-1 accuracy: 70.83 -> 71.27Improved
replace pooling with SoftPool rather than stochastic poolingResNet18 · ImageNet1K val (from scratch, pooling comparison)top-1 accuracy: 70.13 -> 71.27Improved
progressively replace all seven max-pool stages with SoftPool (ablation)InceptionV3 · ImageNet1K val (InceptionV3 progressive replacement)top-1 accuracy: 77.45 -> 79.04Improved
replace only the first pooling layer with SoftPool (ablation)InceptionV3 · ImageNet1K val (InceptionV3 progressive replacement)top-1 accuracy: 77.45 -> 77.93Improved
replace pooling with SoftPoolr3d-50 · HACStop-1 accuracy: 78.36 -> 79.82Improved
replace pooling with SoftPoolSRTG r(2+1)d-50 · HACStop-1 accuracy: 83.77 -> 84.78Improved
replace pooling with SoftPoolSRTG r3d-101 · Kinetics-700 (fine-tuned)top-1 accuracy: 56.46 -> 57.76Improved
replace pooling with SoftPoolSRTG r3d-101 · UCF-101 (fine-tuned)top-1 accuracy: 97.32 -> 98.06Improved

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