AI Paper Atlas

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

Energy-based Self-Training and Normalization for Unsupervised Domain Adaptation

Samitha Herath, Basura Fernando, Ehsan Abbasnejad, Munawar Hayat, Shahram Khadivi, Mehrtash Harandi, Hamid Rezatofighi, Gholamreza Haffari

ICCV 2023 · Label-efficient and robust learning › Domain adaptation and generalization · approach: Feature alignment for unsupervised adaptation · 13 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 free-energy based instance selection to PACMAC consistency selection (joint selection) (ablation)self-training for UDA · DomainNet (average)top-1 accuracy: 80.9 -> 81.5Improved
add free-energy based instance selection to PACMAC consistency selection (joint selection) (ablation)self-training for UDA · OfficeHome (average)top-1 accuracy: 66.8 -> 67.1Improved
use free-energy selection alone (marginal) instead of consistency alone (ablation)self-training for UDA · DomainNet (average)top-1 accuracy: 80.9 -> 80.7No clear effect
select low free-energy instances (proposed) rather than inverted rule, in joint selection (ablation)joint instance selection · DomainNet (average)top-1 accuracy: 80.1 -> 81.5Improved
invert free-energy selection rule in marginal-only case (Sel-high vs Sel-low) (ablation)marginal instance selection · DomainNet (average)top-1 accuracy: 80.7 -> 81.1Mixed
add SCON normalization to free-energy alignment (SCAL) (ablation)SCAL · DomainNet (average)top-1 accuracy: 77.2 -> 82.1Improved
add SCON normalization to free-energy alignment (SCAL) (ablation)SCAL · OfficeHome (average)top-1 accuracy: 63.2 -> 68.2Improved
use SCON normalization alone without SCAL alignment (ablation)SEEBS+ · DomainNet (average)top-1 accuracy: 70.1 -> 46.4Hurt
use proposed SCON loss instead of naive BatchNorm (ablation)SEEBS+ · DomainNet (average)top-1 accuracy: 81.4 -> 83.3Improved
use proposed SCON loss instead of naive BatchNorm (ablation)SEEBS+ · OfficeHome (average)top-1 accuracy: 68.0 -> 68.5Improved
add SCAL and SCON modules to SEEBS (ablation)SEEBS · DomainNet (average)top-1 accuracy: 81.5 -> 83.3Improved
energy-based selection (SEEBS) instead of PACMACViT UDA self-training · VISDA2017accuracy: 77.1 -> 79.1Improved

Earlier methods this paper compares with

Named baselines from earlier accepted papers in this paper's results tables: better than 1 of 1 on most shared results.

Matched by method name in results tables read by AI. Authors' own tables tend to favour the newer method, and a comparison can be missed when a method is renamed.

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