Article
Geochemistry & Geophysics
Xiang Li, Jingyu Deng, Yi Fang
Summary: In this article, a metalearning-based method for few-shot object detection on remote sensing images is introduced. Experimental results on benchmark datasets demonstrate that the proposed method achieves satisfying detection performance with only a few annotated samples and outperforms existing baseline models.
IEEE TRANSACTIONS ON GEOSCIENCE AND REMOTE SENSING
(2022)
Article
Environmental Sciences
Yuteng Ma, Junmin Meng, Baodi Liu, Lina Sun, Hao Zhang, Peng Ren
Summary: In this paper, a novel dictionary learning algorithm is proposed for few-shot remote sensing scene classification. By using natural image datasets for pre-training, a feature extractor suitable for remote sensing data is obtained. A kernel space classifier is designed to map the features to a high-dimensional space and embed the label information into the dictionary learning process to improve feature discrimination for classification. Extensive experiments on four popular remote sensing scene classification datasets demonstrate the effectiveness of the proposed dictionary learning method.
Article
Geochemistry & Geophysics
Zhenqi Cui, Wang Yang, Li Chen, Haifeng Li
Summary: This study proposes a method called metakernel networks (MKNs) to address the challenges in few-shot remote sensing scene classification by integrating a parametric linear classifier (PLC) and metalearning strategy (MKS). Experimental results show a significant improvement in accuracy on three public datasets.
IEEE TRANSACTIONS ON GEOSCIENCE AND REMOTE SENSING
(2022)
Article
Environmental Sciences
Jiayan Wang, Xueqin Wang, Lei Xing, Bao-Di Liu, Zongmin Li
Summary: In recent years, there has been significant attention on few-shot remote sensing scene classification, aiming to achieve excellent performance with limited sample numbers. This study proposes a novel method based on shared class SparsePCA to address the negative transfer problem in few-shot remote sensing scene classification. By using self-supervised learning to assist in training the feature extractor and introducing the Class-Shared SparsePCA classifier, the proposed method significantly improves the classification accuracy on remote sensing datasets.
Article
Geography, Physical
Tianyang Zhang, Xiangrong Zhang, Peng Zhu, Xiuping Jia, Xu Tang, Licheng Jiao
Summary: Recently, there has been increasing attention on few-shot object detection (FSOD) in remote sensing images (RSIs). However, current FSOD methods in RSIs only focus on the detection performance of few-shot novel classes, neglecting the severe degradation of base class performance. This paper proposes a Generalized Few-Shot Detector (G-FSDet) for FSOD in RSIs, which can learn novel knowledge without forgetting and achieves state-of-the-art overall performance.
ISPRS JOURNAL OF PHOTOGRAMMETRY AND REMOTE SENSING
(2023)
Article
Geochemistry & Geophysics
Yuyu Jia, Junyu Gao, Wei Huang, Yuan Yuan, Qi Wang
Summary: This article proposes a holistic mutual representation enhancement method for few-shot segmentation, addressing the issues of intra-class variations and background interference. Extensive experiments demonstrate the superiority of the proposed method and a corresponding benchmark dataset is created.
IEEE TRANSACTIONS ON GEOSCIENCE AND REMOTE SENSING
(2023)
Article
Geochemistry & Geophysics
Xufeng Jiang, Nan Zhou, Xiang Li
Summary: This letter proposes a few-shot learning-based method for semantic segmentation of remote sensing images, which can perform semantic labeling for unseen object categories with limited annotated samples and has good generalization abilities.
IEEE GEOSCIENCE AND REMOTE SENSING LETTERS
(2022)
Article
Geochemistry & Geophysics
Ruolei Li, Yilong Zeng, Jianfeng Wu, Yongli Wang, Xiaoli Zhang
Summary: This paper proposes a two-stage fine-tuning approach for remote sensing images, aiming to solve the issue of category confusion in few-shot object detection. Experimental results demonstrate the effectiveness of the proposed method in various scenarios.
IEEE GEOSCIENCE AND REMOTE SENSING LETTERS
(2022)
Article
Environmental Sciences
Haonan Zhou, Xiaoping Du, Lurui Xia, Sen Li
Summary: In this study, we propose a self-learning method named SFRC for few-shot remote sensing image captioning. SFRC improves the performance in few-shot scenarios by ameliorating the way and efficiency of learning on limited data. The experiments show that SFRC outperforms recent methods in terms of performance evaluation metric scores.
Article
Geochemistry & Geophysics
Jie Chen, Ya Guo, Jingru Zhu, Geng Sun, Dengda Qin, Min Deng, Huimin Liu
Summary: This paper proposes an improved few-shot remote sensing scene classification method that utilizes pretrained word-embedding model to extract semantic information from class names. By fusing textual and visual information through a multimodal prototype fusion module, an enhanced fusion prototype is generated, leading to significant improvement in classification performance.
IEEE TRANSACTIONS ON GEOSCIENCE AND REMOTE SENSING
(2022)
Article
Geochemistry & Geophysics
Lei Wei, Lei Xing, Lifei Zhao, Baodi Liu
Summary: Recently, few-shot scene classification has gained importance in the remote sensing field to address the challenge of insufficient labeled samples. The proposed class-centralized dictionary learning (CCDL) method addresses the weak generalization of training models caused by differing sample distributions between the pretrain and meta-test stages. In the pretraining stage, a model pretrained on a large natural images dataset is fine-tuned using the RS dataset to enhance generalization ability. In the meta-test stage, a CCDL classifier is introduced to ensure distant sparse representations for different categories and concentrated representations for the same category. Experimental results on benchmark datasets demonstrate the effectiveness of the proposed method.
IEEE GEOSCIENCE AND REMOTE SENSING LETTERS
(2023)
Article
Geochemistry & Geophysics
Gong Cheng, Bowei Yan, Peizhen Shi, Ke Li, Xiwen Yao, Lei Guo, Junwei Han
Summary: This article focuses on the main challenges of few-shot object detection in remote sensing images and proposes a simple yet effective method named P-CNN, which consists of a prototype learning network, a prototype-guided region proposal network, and a detector head to overcome the challenges.
IEEE TRANSACTIONS ON GEOSCIENCE AND REMOTE SENSING
(2022)
Article
Geochemistry & Geophysics
Yong Zhou, Han Hu, Jiaqi Zhao, Hancheng Zhu, Rui Yao, Wen-Liang Du
Summary: This letter proposes a few-shot object detection method for the problem of scale variation in remote sensing images. The method includes a context-aware pixel aggregation and a context-aware feature aggregation to adapt to objects at different scales and obtain more semantic information.
IEEE GEOSCIENCE AND REMOTE SENSING LETTERS
(2022)
Article
Geochemistry & Geophysics
Lei Xing, Lifei Zhao, Weijia Cao, Xinmin Ge, Weifeng Liu, Baodi Liu
Summary: This research proposes a class shared dictionary learning (CSDL) approach for few-shot remote sensing scene classification (RSSC). The approach incorporates a mirror-based feature extractor (MFE) and a class shared dictionary (CSD) classifier to enhance classification performance. Extensive experiments validate the effectiveness of the CSDL approach.
IEEE GEOSCIENCE AND REMOTE SENSING LETTERS
(2022)
Article
Geochemistry & Geophysics
Zhong Ji, Liyuan Hou, Xuan Wang, Gang Wang, Yanwei Pang
Summary: This paper proposes a transfer-based dual contrastive network (DCN) to address the challenges of small interclass variances and large intraclass variances in remote sensing image scene classification. The DCN incorporates two auxiliary contrastive learning branches, focusing on interclass discriminability and intraclass invariance. Experimental results on four benchmark datasets demonstrate the competitive performance of the proposed DCN.
IEEE TRANSACTIONS ON GEOSCIENCE AND REMOTE SENSING
(2023)
