We present a method for solving partial differential equations using artificial neural networks and an adaptive collocation strategy. In this procedure, coarse grid of training points is used at the initial stages, while more are added later stages based on value residual larger set evaluation points. This increases robustness network approximation can result in significant computational savings, particularly when solution non-smooth. Numerical results presented benchmark problems...

This paper presents an automatic solution to the problem of detecting and counting cars in unmanned aerial vehicle (UAV) images. is a challenging task given very high spatial resolution UAV images (on order few centimetres) extremely level detail, which require suitable analysis methods. Our proposed method begins by segmenting input image into small homogeneous regions, can be used as candidate locations for car detection. Next, window extracted around each region, deep learning mine highly...

Scene classification is a highly useful task in Remote Sensing (RS) applications. Many efforts have been made to improve the accuracy of RS scene classification. challenging problem, especially for large datasets with tens thousands images number classes and taken under different circumstances. One problem that observed fact given scene, only one part it indicates which class belongs to, whereas other parts are either irrelevant or they actually tend belong another class. To address this...

Abstract We present a stochastic deep collocation method (DCM) based on neural architecture search (NAS) and transfer learning for heterogeneous porous media. first carry out sensitivity analysis to determine the key hyper-parameters of network reduce space subsequently employ hyper-parameter optimization finally obtain parameter values. The presented NAS DCM also saves weights biases most favorable architectures, which is then used in fine-tuning process. techniques drastically...

In this paper, a geometry-based image retrieval system is developed for multi-object images. We model both shape and topology of objects using structured representation called curvature tree (CT). The hierarchy the CT reflects inclusion relationships between objects. To facilitate shape-based matching, triangle-area (TAR) each object stored at corresponding node in CT. similarity two images measured based on maximum subtree isomorphism (MSSI) their CTs. For purpose, we adopt recursive...

In this paper, we present a domain adaptation network to deal with classification scenarios subjected the data shift problem (i.e., labeled and unlabeled images acquired different sensors over completely geographical areas). We rely on power of pretrained convolutional neural networks (CNNs) generate an initial feature representation under analysis, referred as source target domains, respectively. Then feed resulting features extra placed top CNN for further learning. During fine-tuning...

In this article, we propose a novel approach based on convolutional features and sparse autoencoder (AE) for scene-level land-use (LU) classification. This starts by generating an initial feature representation of the scenes under analysis from deep neural network (CNN) pre-learned large amount labelled data auxiliary domain. Then these are fed as input to AE learning new suitable in unsupervised manner. After pre-training phase, two different scenarios building classification system. first...

Recently, a new machine learning approach that is termed as the extreme (ELM) has been introduced in literature. This characterized by unified formulation for regression, binary, and multiclass classification problems, related solution given an analytical compact form. In this letter, we propose efficient method hyperspectral images based on approach. To address model selection issue associated with ELM, develop automatic-solution-based differential evolution (DE). simple yet powerful...

The latest developments in unmanned aerial vehicles (UAVs) and associated sensing systems make these platforms increasingly attractive to the remote community. large amount of spatial details contained images opens door for advanced monitoring applications. In this paper, we use cost-effective technology automatic detection palm trees. Given a UAV image acquired over farm, first extract set keypoints using Scale-invariant Feature Transform (SIFT). Then, analyze with an extreme learning...

We introduce the idea of orthopair membership grades and related general fuzzy sets. It is noted that these generalize intuitionistic Pythagorean sets by allowing support for against to be almost anywhere in [0, 1] × 1], giving systems modelers great freedom capturing human knowledge. The aggregation generalized considered with particular concern OWA Choquet aggregation. concepts possibility certainty as well plausibility belief are investigated this environment. study arithmetic operations...

The current literature of remote sensing (RS) scene classification shows that state-of-the-art results are achieved using feature extraction methods, where convolutional neural networks (CNNs) (mostly VGG16 with 138.36 M parameters) used as extractors and then simple to complex handcrafted modules added for additional learning classification, thus coming back engineering. In this paper, we revisit the fine-tuning approach deeper (GoogLeNet Beyond) show it has not been well exploited due...

Abstract In this work, we present a deep collocation method (DCM) for three-dimensional potential problems in non-homogeneous media. This approach utilizes physics-informed neural network with material transfer learning reducing the solution of partial differential equations to an optimization problem. We tested different configurations including smooth activation functions, sampling methods points generation and combined optimizers. A technique is utilized media gradations parameters, which...

The steady spread of the 2019 Coronavirus disease has brought about human and economic losses, imposing a new lifestyle across world. On this point, medical imaging tests such as computed tomography (CT) X-ray have demonstrated sound screening potential. Deep learning methodologies evidenced superior image analysis capabilities with respect to prior handcrafted counterparts. In paper, we propose novel deep framework for detection using CT images. particular, Vision Transformer architecture...

In this letter, we formulate a land-use (LU) classification problem within compressive sensing (CS) fusion framework. CS aims at providing compact representation form after given query image has been processed with an opportune feature extraction type. particular, residuals are generated from the reconstruction dictionaries associated available set of possible LUs and gathered to single-feature pattern. The patterns obtained different types features then fused provide final LU estimate. Two...