Domain-Invariant Similarity Activation Map Contrastive Learning for Retrieval-Based Long-Term Visual Localization

doi:10.1109/JAS.2021.1003907

CORC > 自动化研究所 > 中国科学院自动化研究所 > 学术期刊 > IEEE/CAA Journal of Automatica Sinica

	Domain-Invariant Similarity Activation Map Contrastive Learning for Retrieval-Based Long-Term Visual Localization
	Hanjiang Hu; Hesheng Wang; Zhe Liu; Weidong Chen
刊名	IEEE/CAA Journal of Automatica Sinica
	2022
卷号	9 期号:2 页码:313-328
关键词	Deep representation learning place recognition visual localization
ISSN号	2329-9266
DOI	10.1109/JAS.2021.1003907
英文摘要	Visual localization is a crucial component in the application of mobile robot and autonomous driving. Image retrieval is an efficient and effective technique in image-based localization methods. Due to the drastic variability of environmental conditions, e.g., illumination changes, retrieval-based visual localization is severely affected and becomes a challenging problem. In this work, a general architecture is first formulated probabilistically to extract domain-invariant features through multi-domain image translation. Then, a novel gradient-weighted similarity activation mapping loss (Grad-SAM) is incorporated for finer localization with high accuracy. We also propose a new adaptive triplet loss to boost the contrastive learning of the embedding in a self-supervised manner. The final coarse-to-fine image retrieval pipeline is implemented as the sequential combination of models with and without Grad-SAM loss. Extensive experiments have been conducted to validate the effectiveness of the proposed approach on the CMU-Seasons dataset. The strong generalization ability of our approach is verified with the RobotCar dataset using models pre-trained on urban parts of the CMU-Seasons dataset. Our performance is on par with or even outperforms the state-of-the-art image-based localization baselines in medium or high precision, especially under challenging environments with illumination variance, vegetation, and night-time images. Moreover, real-site experiments have been conducted to validate the efficiency and effectiveness of the coarse-to-fine strategy for localization.
内容类型	期刊论文
源URL	[http://ir.ia.ac.cn/handle/173211/45992]
专题	自动化研究所_学术期刊_IEEE/CAA Journal of Automatica Sinica
推荐引用方式 GB/T 7714	Hanjiang Hu,Hesheng Wang,Zhe Liu,et al. Domain-Invariant Similarity Activation Map Contrastive Learning for Retrieval-Based Long-Term Visual Localization[J]. IEEE/CAA Journal of Automatica Sinica,2022,9(2):313-328.
APA	Hanjiang Hu,Hesheng Wang,Zhe Liu,&Weidong Chen.(2022).Domain-Invariant Similarity Activation Map Contrastive Learning for Retrieval-Based Long-Term Visual Localization.IEEE/CAA Journal of Automatica Sinica,9(2),313-328.
MLA	Hanjiang Hu,et al."Domain-Invariant Similarity Activation Map Contrastive Learning for Retrieval-Based Long-Term Visual Localization".IEEE/CAA Journal of Automatica Sinica 9.2(2022):313-328.