Due to the limitations of single-source data, joint classification using multisource remote sensing data has received increasing attention. However, existing methods still have certain shortcomings when faced with feature extraction from and fusion between data. In this article, a method based on multiscale interactive information (MIFNet) for hyperspectral synthetic aperture radar (SAR) image is proposed. First, (MIIE) block designed extract meaningful information. Compared traditional...

Due to rich spectral and spatial information, the combination of hyperspectral multispectral images (MSIs) has been widely used for Earth observation, such as wetland classification. However, mining meaningful features effective fusion multisource remote sensing data are still urgent problems be solved. In this article, graph-feature-enhanced selective assignment network (GSANet) is proposed. On one hand, a graph feature extraction module (GFEM) designed extract topological structure...

Deep-learning-based methods are widely used in multisource remote-sensing image classification, and the improvement their performance confirms effectiveness of deep learning for classification tasks. However, inherent underlying problems deep-learning models still hinder further accuracy. For example, after multiple rounds optimization learning, representation bias classifier accumulated, which prevents network performance. In addition, imbalance fusion information among images also leads to...

In recent years, hyperspectral image (HSI) classification based on generative adversarial networks (GAN) has achieved great progress. GAN-based methods can mitigate the limited training sample dilemma to some extent. However, several studies have pointed out that existing HSI are heavily affected by imbalanced data problem. The discriminator in GAN always contradicts itself and tries associate fake labels minority-class samples, thus impair performance. Another critical issue is mode...

Synthetic aperture radar (SAR) image change detection is a vital yet challenging task in the field of remote sensing analysis. Most previous works adopt self-supervised method which uses pseudolabeled samples to guide subsequent training and testing. However, deep networks commonly require many high-quality for parameter optimization. The noise pseudolabels inevitably affects final performance. To solve problem, we propose graph-based knowledge supplement network (GKSNet). be more specific,...

Remote sensing scene understanding is a highly challenging task, and has gradually emerged as research hotspot in the field of intelligent interpretation remote data. Recently, use convolutional neural networks (CNNs) been proven to be fruitful advancement. However, with emergence visual transformers (ViTs), limitations traditional small kernels directly capturing large receptive have posed significant challenges their dominant role. Additionally, fixed neuron connections between different...

Mixup-based data augmentation has been proven to be beneficial the regularization of models during training, especially in remote-sensing field where training is scarce. However, process augmentation, methods ignore target proportion different inputs and keep linear insertion ratio consistent, which leads response label space even if no effective objects are introduced mixed image due randomness process. Moreover, although some previous works have attempted utilize multimodal interaction...

Benefited from the rapid and sustainable development of synthetic aperture radar (SAR) sensors, change detection SAR images has received increasing attentions over past few years. Existing unsupervised deep learning-based methods have made great efforts to exploit robust feature representations, but they consume much time optimize parameters. Besides, these use clustering obtain pseudolabels for training, pseudolabeled samples often involve errors, which can be considered as "label noise."...

In order to enhance efficiency and accuracy of dynamic hand gesture recognition based on HMM method. we propose a new algorithm combined with the DTW for computation's high complexity method in training stage. The can establish relationship fuzzy closeness degree between algorithm. Meanwhile, it combine template adaptive strategy, so that improving algorithm, timeliness human-robot interaction.

Target speech separation is the process of filtering a certain speaker's voice out mixtures according to additional speaker identity information provided. Recent works have made considerable improvement by processing signals in time domain directly. The majority them take fully overlapped for training. However, since most real-life conversations occur randomly and are sparsely overlapped, we argue that training with different overlap ratio data benefits. To do so, an unavoidable problem...

Synthetic aperture radar (SAR) image change detection is a vital yet challenging task in the field of remote sensing analysis. Most previous works adopt self-supervised method which uses pseudo-labeled samples to guide subsequent training and testing. However, deep networks commonly require many high-quality for parameter optimization. The noise pseudo-labels inevitably affects final performance. To solve problem, we propose Graph-based Knowledge Supplement Network (GKSNet). be more...

Objective Fatal road accidents are statistically rare, posing challenges for accurate estimation through the classic logit model (LM). This study seeks to validate efficacy of a rare events logistic (RELM) in enhancing precision fatal crash estimations. Methods Both LM and RELM were employed examine relationship between pertinent risk factors incidence crashes. Crash-injury datasets sourced from Hillsborough County, Florida served as empirical basis evaluating performance metrics both RELM....

Machine Anomalous Sound Detection is crucial for artificial intelligence automation in the context of fourth industrial revolution. Recent approaches employ self-supervised representation learning, which combines representations extracted through training models on normal and pseudo-anomaly data upstream phase with anomaly detection algorithms downstream phase. Although effective extracting demonstrating some robustness, this approach overlooks complexity intrinsic relevance individual tasks...

Explaining multi-agent systems (MAS) is urgent as these become increasingly prevalent in various applications. Previous work has proveided explanations for the actions or states of agents, yet falls short understanding black-boxed agent's importance within a MAS and overall team strategy. To bridge this gap, we propose EMAI, novel agent-level explanation approach that evaluates individual importance. Inspired by counterfactual reasoning, larger change reward caused randomized action agent...

This paper introduces our approaches for the Mask and Breathing Sub-Challenge in Interspeech COMPARE Challenge 2020. For mask detection task, we train deep convolutional neural networks with filter-bank energies, gender-aware features, speaker-aware features. Support Vector Machines follows as back-end classifiers binary prediction on extracted embeddings. Several data augmentation schemes are used to increase quantity of training improve models' robustness, including speed perturbation,...

Change detection using SAR images has drawn increasing attentions in remote sensing communities. It is important to take advantage of the label information changed and unchanged pixels classification. However, most existing methods ignore fact that training process may get corrupt when noise labels exist. To overcome problem, this paper, we study influence image change detection, introduce a random propagation (RLP) algorithm cleanse noise. The key idea RLP use probability transform matrix...

Digital equipment image retrieval is a typical fine-grained problem, In order to solve the necessary data enhancement methods will be essential. Propose method of deep learning in digital device retrieval. Using preprocessing artificial optimal strategy, enhanced set generated through enhancement(ChannelShuffle, ColorJitter), geometric transformation, noise interference, and feature distortion process augmentation improved. It can not only effectively problem large amount manual labeling...

Joint classification of multisource remote sensing data is a very meaningful yet challenging task. Especially when the actual scene composition complex, spectral confusing categories overlapping and different appearances same object will bring great challenges for classification. In this paper, novel Dual Attention Fusion Network (DAFNet) proposed to solve above problems. Firstly, attention block designed highlight or suppress channel map, so as better distinguish categories. At time, we...