KIT | KIT-Bibliothek | Impressum | Datenschutz

ModSelect: Automatic Modality Selection for Synthetic-to-Real Domain Generalization

Marinov, Zdravko ORCID iD icon 1; Roitberg, Alina 1; Schneider, David 1; Stiefelhagen, Rainer ORCID iD icon 1
1 Institut für Anthropomatik und Robotik (IAR), Karlsruher Institut für Technologie (KIT)

Abstract:

Modality selection is an important step when designing multimodal systems, especially in the case of cross-domain activity recognition as certain modalities are more robust to domain shift than others. However, selecting only the modalities which have a positive contribution requires a systematic approach. We tackle this problem by proposing an unsupervised modality selection method (ModSelect), which does not require any ground-truth labels. We determine the correlation between the predictions of multiple unimodal classifiers and the domain discrepancy between their embeddings. Then, we systematically compute modality selection thresholds, which select only modalities with a high correlation and low domain discrepancy. We show in our experiments that our method ModSelect chooses only modalities with positive contributions and consistently improves the performance on a Synthetic-to-Real domain adaptation benchmark, narrowing the domain gap.


Volltext §
DOI: 10.5445/IR/1000158050
Veröffentlicht am 20.04.2023
Cover der Publikation
Zugehörige Institution(en) am KIT Institut für Anthropomatik und Robotik (IAR)
Publikationstyp Forschungsbericht/Preprint
Publikationsjahr 2022
Sprache Englisch
Identifikator KITopen-ID: 1000158050
Umfang 20 S.
Vorab online veröffentlicht am 19.08.2022
Nachgewiesen in arXiv
Dimensions
Relationen in KITopen
KIT – Die Forschungsuniversität in der Helmholtz-Gemeinschaft
KITopen Landing Page