To have a better understanding and usage of Convolution Neural Networks (CNNs), the visualization interpretation CNNs has attracted increasing attention in recent years. In particular, several Class Activation Mapping (CAM) methods been proposed to discover connection between CNN's decision image regions. spite reasonable visualization, lack clear sufficient theoretical support is main limitation these methods. this paper, we introduce two axioms -- Conservation Sensitivity paradigm CAM...

To date, although numerous methods of Change detection (CD) in remote sensing images have been proposed, accurately identifying changes is still a great challenge, due to the difficulties effectively modeling features from ground objects with different patterns. In this paper, novel CD method based on graph convolutional network (GCN) and multiscale object-based technique proposed for both homogeneous heterogeneous images. First, object-wise high level are obtained through pre-trained U-net...

The retrieval performance of a content-based image (CBIR) system is mainly influenced by the feature representations and similarity measures. Recently, deep learning develops rapidly features based on have been applied widely because it has shown that very strong generalization. This paper applies original generated convolution neural network (CNN) to CBIR, uses linear support victor machine (SVM) train hyerplane which can separate similar pairs dissimilar large degree. input SVM in this...

We study the problem of object detection in remote sensing images. As a simple but effective feature extractor, Feature Pyramid Network (FPN) has been widely used several generic vision tasks. However, it still faces some challenges when for detection, as objects images usually exhibit variable shapes, orientations, and sizes. To this end, we propose dedicated detector based on FPN architecture to achieve accurate Specifically, considering shapes orientations objects, first replace original...

In this letter, a novel object detection method based on feature pyramid network (FPN) is proposed to improve the performance of remote sensing objects. First, since information in background regions may interfere with detection, multi-scale deformable attention module (MSDAM) designed and added top backbone FPN make suppress features while highlight target features. The MSDAM generates maps from receptive fields, thus can fit objects various shapes sizes better predict more precise for...

Aims: This paper investigates a nonlinear viscoelastic equation with variable density and memory. Study Design: Polynomial decay. Place Duration of Study: was completed at the School Mathematics Statistics Southwest University during from May 2024 to February 2025. Methodology: Using Faedo-Gal¨erkin method Energy method. Results: We study global well-posedness show polynomial decay results more general weaker assumptions (compared previous studies) on memory kernel. Conclusion: stability,...

Due to the rapid development of deep learning techniques and collection large-scale remote sensing datasets, convolutional neural networks (CNNs) have made significant progress in object detection. However, due diversity objects images, multiscale detection is still a challenging task. In this letter, novel framework based on feature pyramid network (FPN) proposed improve performance objects. First, receptive field expansion block (RFEB) designed added top backbone expand FPN adaptively....

Recently, deep learning algorithms, especially feature pyramid network (FPN), have achieved significant progress in object detection of natural scene images. However, due to the complex scenes remote sensing images and diversity objects, FPN still faces following drawback when applied detection. Specifically, original FPN, features each proposal are extracted by RoIAlign. these limited effective receptive fields, making lack crucial contextual information accurately classify locate as well...

The detection and recognition of dim small infrared (IR) targets across domains pose two formidable challenges: distributional discrepancies in samples scarcity or absence annotated instances the target domain. While current unsupervised domain adaptive object methods can somewhat alleviate performance degradation caused by these issues, they fail to address differences semantic content between background environments different application scenarios. This results a gap that impedes...

Spectra of electron and donor states in quantum dots with different confinement potentials are calculated. The potential-shape, quantum-size, donor-position effects on the level ordering binding studied detail. It is found that a single can heavily change single-electron spectra proper size potential shape, which may be useful for understanding physical phenomena designing materials devices quantum-dot structures.

The very high spatial resolution (VHR) remote sensing images have been an extremely valuable source for monitoring changes occurring on the Earth’s surface. However, precisely detecting relevant in VHR still remains a challenge, due to complexity of relationships among ground objects. To address this limitation, dual neighborhood hypergraph neural network is proposed article, which combines multiscale superpixel segmentation and convolution model exploit complex relationships. First,...

Figure‐ground segmentation is used to extract the foreground from background, where usually defined as region containing most meaningful object of image. In fact, algorithms that take advantage human–computer interaction often attain better performance and they are based on ‘one‐to‐one’ model. this study, authors present a novel algorithm for figure‐ground GrabCut algorithm, which common user interactive. However, instead real user, attempt use pre‐trained deep convolutional neural network...

Interesting features of positron two-dimensional angular correlation annihilation radiation (2D-ACAR) distribution for singly negative divacancies in Si are studied experimentally and theoretically. Anisotropy the is successfully detected a specimen with aligned well reproduced by first-principles calculations based on two-component density-functional theory. The present calculation demonstrates that anisotropy reflects characteristic electrons around divacancies, indicating 2D-ACAR an...

Clustering methods achieve performance improvement by jointly learning representation and cluster assignment. However, they do not consider the confidence of pseudo-labels which are optimal as supervised information, resulting into error accumulation. To address this issue, we propose a Robust Pseudo-labeling for Semantic (RPSC) approach, includes two stages. In first stage (RPSC-Self), design semantic pseudo-labeling scheme using consistency samples, i.e., samples with same semantics should...

The excited states of poly(p-phenylene vinylene) (PPV) are studied in the Su-Schrieffer-Heeger model appended by a long-range Coulomb interaction. We have calculated energy levels singlet and triplet excitons with odd even parity, biexciton state, onset continuum band PPV; these results can comprehensively interpret relevant experimental data. More significantly, we constructed wave function using Heitler-London method binding radius conjugated polymers.

The particle probability hypothesis density (P-PHD) filter gives estimate of target state for multi-target tracking; however, it keeps no record identities and is not able to generate tracks. This paper addresses the problem data association (track continuity) using Density based on cloud aliasing method, that is, corresponding clouds originated from same at two successive time steps overlap each other largely. Thus, suitable associated pairs selected estimated sets can be found tracks step...

Most of the traditional convolution neural network (CNN)‐based classification models are flat classifiers, which have an underlying assumption that all classes equally difficult to distinguish. However, visual separability between different object categories is highly uneven in real world. Recently, hierarchical has been proven effective for CNNs, more and attempts made exploit category hierarchies CNN models. In this study, authors propose a novel architecture, called coarse‐to‐fine CNN. It...

In original data, there may exist redundant features, irrelevant noisy features besides informative features. Extracting while eliminating the others is goal of feature selection. This paper proposed a new selection algorithm based on Relief and SVM-RFE algorithm, it strongly targeted to eliminate unnecessary Finally, We test method three data sets from UCI, treat accuracy, size optimal subset, time-cost as evaluations, experimental results show that has better performance than except time-cost.

Traditional data-driven algorithms suffer from data reliance, hyperparameter sensitivity, and faint characteristics in infrared (IR) "low, slow, small" unmanned aerial vehicle target detection recognition, resulting performance degradation complex backgrounds. Inspired by model-driven methods, this article proposes a learnable feature modulation module that uses prior knowledge to enhance representation. Specifically, method converts the local contrast measure into nonlocal quadrature...

The discriminative optimization (DO) algorithm has been successful in three-dimensional LIDAR rigid point cloud registration. However, its feature descriptor is local, which may restrict the robustness of DO. In this study, for sake improving DO's robustness, we design a global-local descriptor, then weight using prior information from model cloud. We compare proposed approach with eight classical registration methods Stanford Bunny, Oxford SensatUrban and Sydney 3D cross source datasets....

ADVERTISEMENT RETURN TO ISSUEPREVArticleNEXTPhotoinduced Charge Transfer in Excited C60G. P. Zhang, R. T. Fu, X. Sun, K. H. Lee, and Y. ParkCite this: J. Phys. Chem. 1995, 99, 32, 12301–12304Publication Date (Print):August 1, 1995Publication History Published online1 May 2002Published inissue 1 August 1995https://pubs.acs.org/doi/10.1021/j100032a038https://doi.org/10.1021/j100032a038research-articleACS PublicationsRequest reuse permissionsArticle Views98Altmetric-Citations13LEARN ABOUT THESE...