10.1109/WACV51458.2022.00229
https://zenodo.org/records/6363271
oai:zenodo.org:6363271
Victor G. Turrisi da Costa
Victor G. Turrisi da Costa
University of Trento, Italy
Giacomo Zara
Giacomo Zara
University of Trento, Italy
Paolo Rota
Paolo Rota
University of Trento, Italy
Thiago Oliveira-Santos
Thiago Oliveira-Santos
Universidade Federal do Espírito Santo
Niculae Sebe
Niculae Sebe
University of Trento, Italy
Vittorio Murino
Vittorio Murino
University of Verona, Italy
Elisa Ricci
Elisa Ricci
University of Trento, Italy
Dual-Head Contrastive Domain Adaptation for Video Action Recognition
Zenodo
2022
2022-03-16
eng
Creative Commons Attribution 4.0 International
Unsupervised domain adaptation (UDA) methods have become very popular in computer vision. However, while several techniques have been proposed for images, much less attention has been devoted to videos. This paper introduces a novel UDA approach for action recognition from videos, inspired by recent literature on contrastive learning. In particular, we propose a novel two-headed deep architecture that simultaneously adopts cross-entropy and contrastive losses from different network branches to robustly learn a target classifier. Moreover, this work introduces a novel large-scale UDA dataset, Mixamo→Kinetics, which, to the best of our knowledge, is the first dataset that considers the domain shift arising when transferring knowledge from synthetic to real video sequences. Our extensive experimental evaluation conducted on three publicly available benchmarks and on our new Mixamo→Kinetics dataset demonstrate the effectiveness of our approach, which outperforms the current state-of-the-art methods. Code is available at https://github.com/vturrisi/CO2A.
European Commission
10.13039/501100000780
951911
A European Excellence Centre for Media, Society and Democracy