AI Paper Atlas

Research map / Label-efficient and robust learning / Recognition and long-tail

Multi-Scale Spatially-Asymmetric Recalibration for Image Classification

Yan Wang, Lingxi Xie, Siyuan Qiao, Ya Zhang, Wenjun Zhang, Alan L. Yuille

ECCV 2018 · Label-efficient and robust learning › Recognition and long-tail · approach: Loss design and classifier theory · 19 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
add MS-SAR to ResNet-56ResNet-56 · CIFAR10error rate (%): 6.97 -> 6.04Improved
add MS-SAR to ResNet-56ResNet-56 · CIFAR100error rate (%): 29.07 -> 27.71Improved
add MS-SAR to DenseNet-100DenseNet-100 · CIFAR10error rate (%): 4.67 -> 4.16Improved
add MS-SAR to DenseNet-190DenseNet-190 · CIFAR10error rate (%): 3.46 -> 3.32Improved
add MS-SAR to ResNet-20ResNet-20 · CIFAR100error rate (%): 31.87 -> 31.09Improved
add scales 2 and 4 to scale 1 (multi-scale instead of single scale) (ablation)MS-SAR on ResNet-56 · CIFAR10error rate (%): 6.80 -> 6.04Improved
add scale 4 to scale 2 (ablation)MS-SAR on DenseNet-100 · CIFAR10error rate (%): 4.35 -> 4.33 (C10), 21.33 -> 21.23 (C100)Hurt
use regional instead of sliding coordinate sets (ablation)MS-SAR · CIFAR100error rate (%): regional outperforms sliding by ~0.2%Improved
use multi-level instead of single-level recalibration (ablation)MS-SAR on DenseNet-100 · CIFAR10error rate (%): 4.45 -> 4.06Improved
use multi-level instead of single-level recalibration (ablation)MS-SAR on DenseNet-100 · CIFAR100error rate (%): 21.83 -> 21.13Improved
replace SE (single-scale symmetric) with MS-SARResNet-18 · ILSVRC2012 valtop-1 error (%): 29.78 -> 29.43Improved
replace SE with MS-SARResNeXt-50 · ILSVRC2012 valtop-5 error (%): 5.93 -> 5.78Improved

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