AI Paper Atlas

Research map / Label-efficient and robust learning / Test-time adaptation

Deep Nets with Subsampling Layers Unwittingly Discard Useful Activations at Test-Time

Chiao-An Yang, Ziwei Liu, Raymond Yeh

ECCV 2024 · Label-efficient and robust learning › Test-time adaptation · approach: Extrapolation and single-sample adaptation · 1 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 search-and-aggregate of discarded subsampling activations to existing TTA methodsGPS, ClassTTA, AugTTA · ImageNettop-1 accuracy: ClassTTA ResNet18 B=150: 66.40 -> 70.37Improved
add the method on top of TTA baselinesGPS, ClassTTA, AugTTA · Flowers102top-1 accuracy: GPS ResNet18 B=30: 89.04 -> 88.93; AugTTA B=100: 88.86 -> 89.55Mixed
increase total test-time budget for baseline TTA methodsClassTTA · ImageNettop-1 accuracy: ResNet18: 69.09 -> 66.40Hurt
add spatial alignment before aggregation (ablation)learned aggregation · ImageNet (own val split)top-1 accuracy: 79.52 -> 79.88Improved
use learned aggregation instead of simple average (ablation)aggregation · ImageNet (own val split)top-1 accuracy: 79.38 -> 79.88Improved
omit first and last layers from search space (ablation)search · ImageNet (own val split)top-1 accuracy: 79.64 -> 79.88Improved
use learned attention criterion instead of random expansion (ablation)search · ImageNet (own val split)top-1 accuracy: 79.80 -> 79.88Improved
use entropy criterion (training-free) instead of learned criterion (ablation)search · ImageNet (own val split)top-1 accuracy: 79.88 -> 79.86No clear effect
increase B_ours beyond about 15-20 (ablation)search budget · ImageNettop-1 accuracyNo clear effect
add non-learned method to segmentation modelsFCN, DeepLab, SegFormer · CityscapesmIoU: 76.54 -> 77.05 (SegFormer, B=10)Improved
combine method with horizontal flip TTAsegmentation models · ADE20KmIoU: 35.94 -> 36.31 (flip) -> 36.50 (flip + ours)Improved

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