AI Paper Atlas

Research map / Architectures and efficiency / CNN design

SSBNet: Improving Visual Recognition Efficiency by Adaptive Sampling

Ho Man Kwan, Shenghui Song

ECCV 2022 · Architectures and efficiency › CNN design · approach: Efficient CNN scaling and dynamic inference

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
add adaptive instead of uniform sampling (ablation)SSB-ResNet-D-152 (16,8,4), RandAugment · ImageNetTop-1 accuracy: 79.71 -> 80.12Improved
add adaptive instead of uniform sampling (ablation)SSB-ResNet-D-101 (16,8,4), no RandAugment · ImageNetTop-1 accuracy: 78.68 -> 78.88Improved
add adaptive instead of uniform sampling (ablation)SSB-ResNet-D-50 (16,8,4), RandAugment · ImageNetTop-1 accuracy: 78.07 -> 78.22No clear effect
add adaptive instead of uniform sampling (ablation)SSB-ResNet-D-50 (12,6,3), RandAugment · ImageNetTop-1 accuracy: 77.97 -> 77.96No clear effect
use adaptive sampling instead of uniform bilinear sampling (ablation)SSB-ResNet-D-152 (16,8,4) · ImageNetTop-1 accuracy: 78.86 -> 79.05Improved
replace ResNet-D with SSB (adaptive sampling bottleneck)ResNet-D-50 · ImageNetTop-1 accuracy: 78.1 -> 78.1No clear effect
replace ResNet-D with SSB (adaptive sampling bottleneck)ResNet-D-152 · ImageNetTop-1 accuracy: 80.1 -> 79.2Hurt
replace ResNet-RS with SSBResNet-RS-200 · ImageNetTop-1 accuracy: 83.4 -> 82.6Hurt
replace EfficientNet with SSBEfficientNet-B0 · ImageNetTop-1 accuracy: 76.4 -> 75.4Hurt
replace ResNet-D-50 with SSB-ResNet-D-50 backboneMask R-CNN FPN · COCO-2017 (12 epochs)AP box: 40.14 -> 40.67Improved
replace ResNet-D-50 with SSB-ResNet-D-50 backboneMask R-CNN FPN · COCO-2017 (36 epochs)AP box: 42.50 -> 42.89Improved
replace ResNet-RS-200 with SSB versionResNet-RS-200 · Latency V100latency: 988 -> 744Improved

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