AI Paper Atlas

Research map / Label-efficient and robust learning / Domain adaptation and generalization

Tune It the Right Way: Unsupervised Validation of Domain Adaptation via Soft Neighborhood Density

Kuniaki Saito, Donghyun Kim, Piotr Teterwak, Stan Sclaroff, Trevor Darrell, Kate Saenko

ICCV 2021 · Label-efficient and robust learning › Domain adaptation and generalization · approach: Feature alignment for unsupervised adaptation · 45 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
select checkpoints with SND instead of Source Riskunsupervised DA validation · Closed DA Office+OfficeHometarget accuracy (%): 71.1 -> 74.3Improved
select with SND instead of Entropyunsupervised DA validation · Closed DA Office+OfficeHometarget accuracy (%): 71.9 -> 74.3Improved
select with SND instead of DEVunsupervised DA validation · Closed DA Office+OfficeHometarget accuracy (%): 71.6 -> 74.3Improved
select with SND instead of Entropyunsupervised DA validation · Partial DA OfficeHometarget accuracy (%): 60.1 -> 66.0Improved
select with SND instead of Source Riskunsupervised DA validation · Partial DA OfficeHometarget accuracy (%): 62.7 -> 66.0Improved
select with SND instead of Entropyunsupervised DA validation · VisDA and DomainNet closed DAtarget accuracy (%): 64.9 -> 65.8No clear effect
select with SND instead of Entropy on ADVENTunsupervised DA validation · GTA5 to CityscapesmIoU: 35.6 -> 40.2Improved
select with SND instead of Entropy on AdaptSegunsupervised DA validation · GTA5 to CityscapesmIoU: 39.3 -> 39.5No clear effect
use last-layer softmax output as input feature instead of middle-layer features (ablation)SND · Office A to Dwithin-class variance: 0.437 -> 0.163Improved
add softmax to last-layer output as input feature (ablation)SND · Office A to Dwithin-class variance: 0.215 -> 0.163Improved
add softmax to last-layer output as input feature (ablation)SND · OfficeHome A to Pwithin-class variance: 0.301 -> 0.285Improved
use SND to choose source domainSND · OfficeHome source domain selection, Target Realaccuracy (%): 74.6 -> 76.0Mixed

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