Cross-dataset Learning for Generalizable Land Use Scene Classification
Research output: Chapter in Book/Report/Conference proceeding › Article in proceedings › Research › peer-review
Standard
Cross-dataset Learning for Generalizable Land Use Scene Classification. / Gominski, Dimitri; Gouet-Brunet, Valerie; Chen, Liming.
Proceedings - 2022 IEEE/CVF Conference on Computer Vision and Pattern Recognition Workshops, CVPRW 2022. IEEE Computer Society Press, 2022. p. 1381-1390 (IEEE Computer Society Conference on Computer Vision and Pattern Recognition Workshops, Vol. 2022-June).Research output: Chapter in Book/Report/Conference proceeding › Article in proceedings › Research › peer-review
Harvard
APA
Vancouver
Author
Bibtex
}
RIS
TY - GEN
T1 - Cross-dataset Learning for Generalizable Land Use Scene Classification
AU - Gominski, Dimitri
AU - Gouet-Brunet, Valerie
AU - Chen, Liming
N1 - Publisher Copyright: © 2022 IEEE.
PY - 2022
Y1 - 2022
N2 - Few-shot and cross-domain land use scene classification methods propose solutions to classify unseen classes or un-seen visual distributions, but are hardly applicable to real-world situations due to restrictive assumptions. Few-shot methods involve episodic training on restrictive training subsets with small feature extractors, while cross-domain methods are only applied to common classes. The underlying challenge remains open: can we accurately classify new scenes on new datasets? In this paper, we propose a new framework for few-shot, cross-domain classification. Our retrieval-inspired approach1 exploits the interrelations in both the training and testing data to output class labels using compact descriptors. Results show that our method can accurately produce land-use predictions on unseen datasets and unseen classes, going beyond the traditional few-shot or cross-domain formulation, and allowing cross-dataset training.
AB - Few-shot and cross-domain land use scene classification methods propose solutions to classify unseen classes or un-seen visual distributions, but are hardly applicable to real-world situations due to restrictive assumptions. Few-shot methods involve episodic training on restrictive training subsets with small feature extractors, while cross-domain methods are only applied to common classes. The underlying challenge remains open: can we accurately classify new scenes on new datasets? In this paper, we propose a new framework for few-shot, cross-domain classification. Our retrieval-inspired approach1 exploits the interrelations in both the training and testing data to output class labels using compact descriptors. Results show that our method can accurately produce land-use predictions on unseen datasets and unseen classes, going beyond the traditional few-shot or cross-domain formulation, and allowing cross-dataset training.
U2 - 10.1109/CVPRW56347.2022.00144
DO - 10.1109/CVPRW56347.2022.00144
M3 - Article in proceedings
AN - SCOPUS:85137790889
T3 - IEEE Computer Society Conference on Computer Vision and Pattern Recognition Workshops
SP - 1381
EP - 1390
BT - Proceedings - 2022 IEEE/CVF Conference on Computer Vision and Pattern Recognition Workshops, CVPRW 2022
PB - IEEE Computer Society Press
T2 - 2022 IEEE/CVF Conference on Computer Vision and Pattern Recognition Workshops, CVPRW 2022
Y2 - 19 June 2022 through 20 June 2022
ER -
ID: 344438256