Domain Decorrelation with Potential Energy Ranking

Pei, S; Sun, J; Xu, RYD; Xiang, S; Meng, G

Domain Decorrelation with Potential Energy Ranking

Pei, S Sun, J Xu, RYD Xiang, S Meng, G

Permalink

Publisher:: Association for the Advancement of Artificial Intelligence (AAAI)
Publication Type:: Conference Proceeding
Citation:: Proceedings of the 37th AAAI Conference on Artificial Intelligence, AAAI 2023, 2023, 37, (2), pp. 2020-2028
Issue Date:: 2023-06-27

Recently Added

	Filename	Description	Size
	25294-Article Text-29357-1-2-20230626.pdf	Published version	8.42 MB	Adobe PDF	View/Open

Copyright Clearance Process

Recently Added
In Progress
Open Access

This item is new to OPUS and is not currently available.

Full metadata record

Field	Value	Language
dc.contributor.author	Pei, S
dc.contributor.author	Sun, J
dc.contributor.author	Xu, RYD
dc.contributor.author	Xiang, S
dc.contributor.author	Meng, G
dc.date.accessioned	2024-05-10T10:37:20Z
dc.date.available	2024-05-10T10:37:20Z
dc.date.issued	2023-06-27
dc.identifier.citation	Proceedings of the 37th AAAI Conference on Artificial Intelligence, AAAI 2023, 2023, 37, (2), pp. 2020-2028
dc.identifier.isbn	9781577358800
dc.identifier.issn	2159-5399
dc.identifier.issn	2374-3468
dc.identifier.uri	http://hdl.handle.net/10453/178808
dc.description.abstract	Machine learning systems, especially the methods based on deep learning, enjoy great success in modern computer vision tasks under ideal experimental settings. Generally, these classic deep learning methods are built on the i.i.d. assumption, supposing the training and test data are drawn from the same distribution independently and identically. However, the aforementioned i.i.d. assumption is, in general, unavailable in the real-world scenarios, and as a result, leads to sharp performance decay of deep learning algorithms. Behind this, domain shift is one of the primary factors to be blamed. In order to tackle this problem, we propose using Potential Energy Ranking (PoER) to decouple the object feature and the domain feature in given images, promoting the learning of label-discriminative representations while filtering out the irrelevant correlations between the objects and the background. PoER employs the ranking loss in shallow layers to make features with identical category and domain labels close to each other and vice versa. This makes the neural networks aware of both objects and background characteristics, which is vital for generating domain-invariant features. Subsequently, with the stacked convolutional blocks, PoER further uses the contrastive loss to make features within the same categories distribute densely no matter domains, filtering out the domain information progressively for feature alignment. PoER reports superior performance on domain generalization benchmarks, improving the average top-1 accuracy by at least 1.20% compared to the existing methods. Moreover, we use PoER in the ECCV 2022 NICO Challenge, achieving top place with only a vanilla ResNet-18 and winning the jury award. The code has been made publicly available at: https://github.com/ForeverPs/PoER.
dc.language	en
dc.publisher	Association for the Advancement of Artificial Intelligence (AAAI)
dc.relation.ispartof	Proceedings of the 37th AAAI Conference on Artificial Intelligence, AAAI 2023
dc.relation.isbasedon	10.1609/aaai.v37i2.25294
dc.rights	info:eu-repo/semantics/restrictedAccess
dc.title	Domain Decorrelation with Potential Energy Ranking
dc.type	Conference Proceeding
utslib.citation.volume	37
pubs.organisational-group	University of Technology Sydney
pubs.organisational-group	University of Technology Sydney/Faculty of Engineering and Information Technology
pubs.organisational-group	University of Technology Sydney/Strength - INEXT - Innovation in IT Services and Applications
pubs.organisational-group	University of Technology Sydney/Strength - GBDTC - Global Big Data Technologies
pubs.organisational-group	University of Technology Sydney/Faculty of Engineering and Information Technology/School of Electrical and Data Engineering
utslib.copyright.status	recently_added	*
dc.date.updated	2024-05-10T10:37:16Z
pubs.issue	2
pubs.publication-status	Published
pubs.volume	37
utslib.citation.issue	2

Abstract:

Machine learning systems, especially the methods based on deep learning, enjoy great success in modern computer vision tasks under ideal experimental settings. Generally, these classic deep learning methods are built on the i.i.d. assumption, supposing the training and test data are drawn from the same distribution independently and identically. However, the aforementioned i.i.d. assumption is, in general, unavailable in the real-world scenarios, and as a result, leads to sharp performance decay of deep learning algorithms. Behind this, domain shift is one of the primary factors to be blamed. In order to tackle this problem, we propose using Potential Energy Ranking (PoER) to decouple the object feature and the domain feature in given images, promoting the learning of label-discriminative representations while filtering out the irrelevant correlations between the objects and the background. PoER employs the ranking loss in shallow layers to make features with identical category and domain labels close to each other and vice versa. This makes the neural networks aware of both objects and background characteristics, which is vital for generating domain-invariant features. Subsequently, with the stacked convolutional blocks, PoER further uses the contrastive loss to make features within the same categories distribute densely no matter domains, filtering out the domain information progressively for feature alignment. PoER reports superior performance on domain generalization benchmarks, improving the average top-1 accuracy by at least 1.20% compared to the existing methods. Moreover, we use PoER in the ECCV 2022 NICO Challenge, achieving top place with only a vanilla ResNet-18 and winning the jury award. The code has been made publicly available at: https://github.com/ForeverPs/PoER.

Please use this identifier to cite or link to this item:

http://hdl.handle.net/10453/178808