AI Paper Atlas

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

Open Set Domain Adaptation for Image and Action Recognition

Pau Panareda Busto, Ahsan Iqbal, Jüergen Gall

TPAMI 2018 · Label-efficient and robust learning › Domain adaptation and generalization · approach: Source-free domain adaptation · 101 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 assign-and-transform-iteratively domain adaptation (ATI) over source-only LSVMLSVM · Office OS (unsup, all samples)accuracy: 68.3 -> 82.6Improved
add outlier rejection (lambda) to ATI (ablation)ATI · Office OS (unsup, 5 splits)accuracy: 81.5 -> 82.3Improved
add outlier rejection (lambda) to ATI (ablation)ATI · Office CS (unsup, all samples)accuracy: 93.8 -> 93.7No clear effect
add locality-constrained assignment with 1 neighbour to ATI-λ in unsupervised setting (ablation)ATI-λ · Office OS (unsup, all samples)accuracy: 82.9 -> 81.3Hurt
add locality-constrained assignment with 1 neighbour to ATI-λ in semi-supervised setting (ablation)ATI-λ · Office OS (semi-sup)accuracy: 85.7 -> 86.0Improved
include the unknown class in the label set C (ablation)ATI-λ · Office OS (unsup, all samples)accuracy: 82.3 -> 82.9Improved
use open set SVM instead of linear SVM on transformed source (ablation)ATI-λ · Office OS (unsup, all samples)accuracy: 82.9 -> 75.0Hurt
increase number of unknown samples from 20 to 400 (ablation)LSVM · Amazon->DSLR+Webcam OSaccuracy: 74.2 -> 48.2Hurt
increase number of unknown samples from 20 to 400 (ablation)ATI-λ · Amazon->DSLR+Webcam OSaccuracy: 80.3 -> 81.9No clear effect
lower rho towards 0 (more outlier rejection) (ablation)ATI-λ · Office OS (unsup)accuracyHurt
add constraint that every class gets at least one assigned target sample (ablation)ATI-λ · Office OS (unsup)accuracyImproved
use VGG-16 features instead of AlexNetATI-λ · VISDA->Pascal3D OSaccuracy: 61.4 -> 71.9Improved

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