Related references
Note: Only part of the references are listed.
Article
Computer Science, Artificial Intelligence
Minjia Yang et al.
Summary: Restricted by observation conditions, effective classification of scarce targets in SAR image is challenging due to few samples. Existing few-shot SAR target classification methods have made breakthroughs but ignore part-level features, leading to degraded performance in fine-grained classification. To address this issue, a novel few-shot fine-grained classification framework called HENC is proposed, which extracts multi-scale features from both object-level and part-level using hierarchical embedding network (HEN) and realizes joint inference of these features with scale-channels. Moreover, a center calibration algorithm is proposed to explicitly calibrate novel centers by utilizing center information of base categories, improving classification accuracy for SAR targets.
IEEE TRANSACTIONS ON IMAGE PROCESSING
(2023)
Article
Geochemistry & Geophysics
Gong Cheng et al.
Summary: SPNet is a lightweight and effective few-shot image classification model, utilizing prototype self-calibration and intercalibration methods to generate more accurate prototypes and predictions through optimizing three loss functions, demonstrating competitive performance compared with other advanced few-shot image classification approaches.
IEEE TRANSACTIONS ON GEOSCIENCE AND REMOTE SENSING
(2022)
Article
Geochemistry & Geophysics
Minjia Yang et al.
Summary: In this paper, a few-shot learning (FSL) framework based on deep learning is proposed to address the problem of insufficient training samples in synthetic aperture radar (SAR) automatic target classification. The framework includes stages of data augmentation, embedding network, multilayer graph attention network, and mixed loss, which can improve classification accuracy and robustness.
IEEE TRANSACTIONS ON GEOSCIENCE AND REMOTE SENSING
(2022)
Proceedings Paper
Computer Science, Artificial Intelligence
Haoxing Chen et al.
Summary: This paper proposes a novel Multi-Scale Adaptive Task Attention Network (MATANet) for few-shot learning. By constructing a multi-scale feature generator, an adaptive task attention module, and a similarity-to-class module, it addresses the limitations of existing methods in handling each category independently and the coexistence of dominant objects at different scales.
2022 26TH INTERNATIONAL CONFERENCE ON PATTERN RECOGNITION (ICPR)
(2022)
Proceedings Paper
Computer Science, Artificial Intelligence
Jiangtao Xie et al.
Summary: This paper proposes a deep Brownian Distance Covariance (DeepBDC) method for few-shot classification. The central idea is to learn image representations by measuring the discrepancy between joint characteristic functions of embedded features and product of marginals. Extensive evaluations show that DeepBDC significantly outperforms other methods and achieves state-of-the-art results.
2022 IEEE/CVF CONFERENCE ON COMPUTER VISION AND PATTERN RECOGNITION (CVPR)
(2022)
Article
Geochemistry & Geophysics
Yangfan Li et al.
Summary: Fine-grained ship classification is important but challenging in machine learning due to limited availability of ship images. This study proposes a foreground-aware feature map reconstruction network (FRN) that achieves state-of-the-art results on ship classification datasets.
IEEE TRANSACTIONS ON GEOSCIENCE AND REMOTE SENSING
(2022)
Article
Geochemistry & Geophysics
Li Wang et al.
Summary: In this article, a novel few-shot learning framework named hybrid inference network (HIN) is proposed to address the problem of SAR target recognition with only a few training samples. The recognition procedure of HIN consists of two main stages: embedding the SAR images into an embedding space using an embedding network and classifying the samples in the embedding space using a hybrid inference strategy that combines inductive and transductive inference methods. The experimental results show that HIN performs well in few-shot SAR image classification.
IEEE TRANSACTIONS ON GEOSCIENCE AND REMOTE SENSING
(2021)
Article
Environmental Sciences
Yanghua Di et al.
Summary: The paper investigates fine-grained ship classification in remote sensing images, determines 42 common categories, and creates the FGSCR-42 dataset. Researchers collect remote sensing images containing warships and civilian ships of various scales from multiple datasets, and make the dataset publicly available for use.
Proceedings Paper
Computer Science, Artificial Intelligence
Dahyun Kang et al.
Summary: This method proposes to address few-shot classification by meta-learning what to observe and where to attend in a relational perspective. By leveraging self-correlational representation and cross-correlational attention modules, it consistently improves state-of-the-art methods on widely used few-shot classification benchmarks in experimental evaluation.
2021 IEEE/CVF INTERNATIONAL CONFERENCE ON COMPUTER VISION (ICCV 2021)
(2021)
Proceedings Paper
Acoustics
Qing-Long Zhang et al.
Summary: Attention mechanisms are crucial in enhancing the performance of deep neural networks, with the Shuffle Attention (SA) module effectively combining two types of attention mechanisms to achieve better performance while reducing computational complexity.
2021 IEEE INTERNATIONAL CONFERENCE ON ACOUSTICS, SPEECH AND SIGNAL PROCESSING (ICASSP 2021)
(2021)
Article
Engineering, Electrical & Electronic
Jiawei Shi et al.
Summary: This study proposes a metric-based few-shot learning method to generate novel concept representation using nearest neighbor prototype, aiming to address the degradation of CNN classification performance in the case of few training samples. By mapping samples to the feature space through CNN and generating prototypes by computing nearest neighbor value on each dimension of the feature separately, the accuracy and robustness of ship classification with a small amount of labeled data are verified.
IEEE JOURNAL OF SELECTED TOPICS IN APPLIED EARTH OBSERVATIONS AND REMOTE SENSING
(2021)
Article
Computer Science, Information Systems
Zhengwu Yuan et al.
Summary: This paper proposes an efficient few-shot scene classification scheme in remote sensing by combining multiple attention mechanisms and the attention-reference mechanism into the deepEMD network. The experimental results show that the proposed scheme achieves state-of-the-art results in few-shot remote sensing scene classification by capturing both global and local discriminative information.
Article
Computer Science, Artificial Intelligence
Xiaoxu Li et al.
Summary: The proposed Bi-Similarity Network (BSNet) utilizes two similarity measures to learn more discriminative and less similarity-biased features from few shots of fine-grained images, significantly improving model generalization ability. Extensive experiments show substantial improvement on several fine-grained image benchmark datasets, demonstrating the effectiveness of the approach. The codes for the model are available at: https://github.com/PRIS-CV/BSNet.
IEEE TRANSACTIONS ON IMAGE PROCESSING
(2021)
Proceedings Paper
Computer Science, Artificial Intelligence
Yongqin Xian et al.
2019 IEEE/CVF CONFERENCE ON COMPUTER VISION AND PATTERN RECOGNITION (CVPR 2019)
(2019)
Proceedings Paper
Computer Science, Artificial Intelligence
Muhammad Abdullah Jamal et al.
2019 IEEE/CVF CONFERENCE ON COMPUTER VISION AND PATTERN RECOGNITION (CVPR 2019)
(2019)
Proceedings Paper
Computer Science, Artificial Intelligence
Hongguang Zhang et al.
2019 IEEE/CVF CONFERENCE ON COMPUTER VISION AND PATTERN RECOGNITION (CVPR 2019)
(2019)
Proceedings Paper
Computer Science, Artificial Intelligence
Amit Alfassy et al.
2019 IEEE/CVF CONFERENCE ON COMPUTER VISION AND PATTERN RECOGNITION (CVPR 2019)
(2019)
Proceedings Paper
Computer Science, Artificial Intelligence
Yu-Xiong Wang et al.
2018 IEEE/CVF CONFERENCE ON COMPUTER VISION AND PATTERN RECOGNITION (CVPR)
(2018)
Article
Statistics & Probability
Gabor J. Szekely et al.
ANNALS OF STATISTICS
(2007)