Article
Geochemistry & Geophysics
Danfeng Hong, Zhu Han, Jing Yao, Lianru Gao, Bing Zhang, Antonio Plaza, Jocelyn Chanussot
Summary: The article introduces a novel HS image classification network called SpectralFormer, which utilizes the transformers framework to learn spectral sequence information, achieving better classification performance than traditional methods and exhibiting high flexibility.
IEEE TRANSACTIONS ON GEOSCIENCE AND REMOTE SENSING
(2022)
Proceedings Paper
Computer Science, Artificial Intelligence
Ali Hatamizadeh, Yucheng Tang, Vishwesh Nath, Dong Yang, Andriy Myronenko, Bennett Landman, Holger R. Roth, Daguang Xu
Summary: Fully Convolutional Neural Networks (FCNNs) have been successful in medical image segmentation, but their limited ability to learn long-range dependencies is a challenge. Inspired by transformers in NLP, we propose a novel architecture called UNet Transformers (UNETR) to redefine volumetric medical image segmentation as a sequence prediction problem. By combining transformers and U-shaped network design in the encoder and decoder, we effectively capture global information and achieve semantic segmentation output.
2022 IEEE WINTER CONFERENCE ON APPLICATIONS OF COMPUTER VISION (WACV 2022)
(2022)
Article
Geochemistry & Geophysics
Hao Chen, Zipeng Qi, Zhenwei Shi
Summary: This study introduces a bitemporal image transformer (BIT) for efficient and effective change detection by modeling contexts in the spatial-temporal domain. The BIT model demonstrates superior performance and efficiency on three CD datasets, significantly outperforming the purely convolutional baseline model with lower computational costs.
IEEE TRANSACTIONS ON GEOSCIENCE AND REMOTE SENSING
(2022)
Proceedings Paper
Computer Science, Artificial Intelligence
Kaiming He, Xinlei Chen, Saining Xie, Yanghao Li, Piotr Dollar, Ross Girshick
Summary: This paper presents a self-supervised learning method for computer vision based on masked autoencoders. By masking a portion of the input image and reconstructing the missing pixels, large models can be trained efficiently and effectively. The approach achieves high generalization performance and outperforms supervised pretraining in transfer learning tasks.
2022 IEEE/CVF CONFERENCE ON COMPUTER VISION AND PATTERN RECOGNITION (CVPR 2022)
(2022)
Article
Geochemistry & Geophysics
Jiaming Han, Jian Ding, Jie Li, Gui-Song Xia
Summary: Significant progress has been made in the past decade on detecting objects in aerial images. We propose a single-shot alignment network (S(2)A-Net) that consists of two modules to address the misalignment issue between anchors and convolutional features, improving the consistency between classification score and localization accuracy.
IEEE TRANSACTIONS ON GEOSCIENCE AND REMOTE SENSING
(2022)
Proceedings Paper
Computer Science, Artificial Intelligence
Zhuang Liu, Hanzi Mao, Chao-Yuan Wu, Christoph Feichtenhofer, Trevor Darrell, Saining Xie
Summary: The development of visual recognition has gone through stages from ConvNets to ViTs and then to hybrid approaches. In this work, the design of a pure ConvNet is reexamined and several key components are discovered, resulting in the construction of the ConvNeXt model series. These models compete with Transformers in terms of accuracy and performance while maintaining the simplicity and efficiency of ConvNets.
2022 IEEE/CVF CONFERENCE ON COMPUTER VISION AND PATTERN RECOGNITION (CVPR)
(2022)
Article
Geochemistry & Geophysics
Sheng Fang, Kaiyu Li, Jinyuan Shao, Zhe Li
Summary: This letter proposes a densely connected siamese network (SNUNet-CD) for change detection, which alleviates the loss of localization information in deep layers and introduces ECAM for deep supervision. Experimental results show that the method achieves a better tradeoff between accuracy and calculation amount compared to other state-of-the-art change detection methods.
IEEE GEOSCIENCE AND REMOTE SENSING LETTERS
(2022)
Article
Geochemistry & Geophysics
Le Sun, Guangrui Zhao, Yuhui Zheng, Zebin Wu
Summary: In this article, the spectral-spatial feature tokenization transformer (SSFTT) method is proposed to capture spectral-spatial and high-level semantic features. Experimental analysis confirms that this method outperforms other deep learning methods in terms of computation time and classification performance.
IEEE TRANSACTIONS ON GEOSCIENCE AND REMOTE SENSING
(2022)
Proceedings Paper
Computer Science, Artificial Intelligence
Robin Rombach, Andreas Blattmann, Dominik Lorenz, Patrick Esser, Bjoern Ommer
Summary: By applying diffusion models (DMs) in the latent space of powerful pretrained autoencoders, this paper achieves high-quality image generation on limited computational resources. The introduction of cross-attention layers further enhances the flexibility and performance of the generator.
2022 IEEE/CVF CONFERENCE ON COMPUTER VISION AND PATTERN RECOGNITION (CVPR)
(2022)
Article
Geochemistry & Geophysics
Xintong Dong, Yue Li, Tie Zhong, Ning Wu, Hongzhou Wang
Summary: Distributed fiber-optical acoustic sensing (DAS) is a promising technology in seismic exploration, but the quality of DAS-VSP data is often affected by random and coherent noises. To improve the signal-to-noise ratio, a CNN model based on L-FM-CNN is proposed, utilizing leaky ReLU as the activation function. By constructing a high-authenticity theoretical seismic data set and using a new loss function ERM, the proposed method proves effective in denoising DAS-VSP data with different SNRs.
IEEE GEOSCIENCE AND REMOTE SENSING LETTERS
(2022)
Article
Geochemistry & Geophysics
Weiwei Cai, Zhanguo Wei
Summary: In this letter, a novel cross-attention mechanism and graph convolution integration algorithm are proposed for hyperspectral data classification. Experimental results show that the proposed algorithm achieves better performances than other algorithms using different methods of training set division.
IEEE GEOSCIENCE AND REMOTE SENSING LETTERS
(2022)
Article
Geochemistry & Geophysics
Xin Wu, Danfeng Hong, Jocelyn Chanussot
Summary: This paper proposes a new framework for multimodal remote sensing data classification, using deep learning and a cross-channel reconstruction module to learn compact fusion representations of different data sources. Extensive experiments on two multimodal RS datasets demonstrate the effectiveness and superiority of the proposed method.
IEEE TRANSACTIONS ON GEOSCIENCE AND REMOTE SENSING
(2022)
Article
Geochemistry & Geophysics
Qian Shi, Mengxi Liu, Shengchen Li, Xiaoping Liu, Fei Wang, Liangpei Zhang
Summary: The article proposes a deeply supervised attention metric-based network (DSAMNet) to address the challenges in change detection. The network uses a metric module for deep metric learning, integrated with convolutional block attention modules (CBAM), and a DS module to enhance feature extraction and generate more useful features. A new CD dataset, Sun Yat-Sen University (SYSU)-CD, containing 20,000 aerial image pairs, is also created for bitemporal image CD. Experimental results show that the network achieves the highest accuracy on both datasets.
IEEE TRANSACTIONS ON GEOSCIENCE AND REMOTE SENSING
(2022)
Proceedings Paper
Computer Science, Artificial Intelligence
Zhendong Wang, Xiaodong Cun, Jianmin Bao, Wengang Zhou, Jianzhuang Liu, Houqiang Li
Summary: This paper introduces Uformer, an image restoration architecture based on Transformer, with a hierarchical encoder-decoder network and novel designs including locally-enhanced window Transformer block and learnable multi-scale restoration modulator. Uformer demonstrates high capability for image restoration tasks and achieves superior performance in various experiments.
2022 IEEE/CVF CONFERENCE ON COMPUTER VISION AND PATTERN RECOGNITION (CVPR 2022)
(2022)
Article
Computer Science, Interdisciplinary Applications
Richard J. Chen, Ming Y. Lu, Jingwen Wang, Drew F. K. Williamson, Scott J. Rodig, Neal Lindeman, Faisal Mahmood
Summary: This study proposes an interpretable strategy for multimodal fusion of histology image and genomic features for survival outcome prediction. The results on glioma and clear cell renal cell carcinoma datasets demonstrate that this approach improves the prognostic determinations.
IEEE TRANSACTIONS ON MEDICAL IMAGING
(2022)
Article
Geochemistry & Geophysics
Fulin Luo, Zehua Zou, Jiamin Liu, Zhiping Lin
Summary: The research proposes a multistructure unified discriminative embedding (MUDE) method to extract the low-dimensional features of hyperspectral image (HSI), by considering the neighborhood, tangential, and statistical properties of each sample in HSI for achieving the complementarity of different characteristics. Experimental results demonstrate that the proposed method can improve the classification performance of HSI.
IEEE TRANSACTIONS ON GEOSCIENCE AND REMOTE SENSING
(2022)
Article
Environmental Sciences
Jing Wei, Zhanqing Li, Ke Li, Russell R. Dickerson, Rachel T. Pinker, Jun Wang, Xiong Liu, Lin Sun, Wenhao Xue, Maureen Cribb
Summary: This study presents a new method to estimate ground-level ozone concentrations in China using solar radiation intensity and surface temperature. The generated dataset provides reliable and valuable information, including short-term severe ozone pollution events and the rapid increase and recovery of ozone concentrations associated with changes in anthropogenic emissions. The study also reveals an increase in summertime ozone concentrations and the probability of daily ozone pollution since 2015, with a decline observed in 2020 due to air pollution control measures and the COVID-19 pandemic.
REMOTE SENSING OF ENVIRONMENT
(2022)
Review
Geography, Physical
Xian Sun, Peijin Wang, Zhiyuan Yan, Feng Xu, Ruiping Wang, Wenhui Diao, Jin Chen, Jihao Li, Yingchao Feng, Tao Xu, Martin Weinmann, Stefan Hinz, Cheng Wang, Kun Fu
Summary: With the rapid development of deep learning, many deep learning-based approaches have achieved great success in object detection tasks. However, existing datasets have limitations in terms of scale, category, and image. To address the needs of high-resolution remote sensing images, we propose a novel benchmark dataset called FAIR1M, which includes over 1 million instances and more than 40,000 images for fine-grained object recognition. The FAIR1M dataset has several unique characteristics, such as being larger than other datasets, providing richer category information, containing geographic information, and having better image quality. We evaluate state-of-the-art methods on the FAIR1M dataset and propose improvements to the evaluation metrics and the incorporation of hierarchy detection. We believe that the FAIR1M dataset will contribute to the field of earth observation through fine-grained object detection in large-scale real-world scenes.
ISPRS JOURNAL OF PHOTOGRAMMETRY AND REMOTE SENSING
(2022)
Article
Geochemistry & Geophysics
Rui Li, Shunyi Zheng, Ce Zhang, Chenxi Duan, Jianlin Su, Libo Wang, Peter M. Atkinson
Summary: Semantic segmentation of remote sensing images is vital for various applications such as land resource management and urban planning. Despite the improvement in accuracy with deep convolutional neural networks, standard models have limitations like underuse of information and insufficient exploration of long-range dependencies. This article introduces a multiattention network (MANet) with efficient attention modules to address these issues and demonstrates superior performance on large-scale remote sensing datasets.
IEEE TRANSACTIONS ON GEOSCIENCE AND REMOTE SENSING
(2022)
Article
Geochemistry & Geophysics
Yao Ding, Xiaofeng Zhao, Zhili Zhang, Wei Cai, Nengjun Yang, Ying Zhan
Summary: A novel dense graph neural network structure incorporating ARMA filters, dense structure, and context-aware learning mechanism has been proposed and applied successfully to hyperspectral image classification. Experimental results demonstrated its superiority over current state-of-the-art methods.
IEEE TRANSACTIONS ON GEOSCIENCE AND REMOTE SENSING
(2022)
