Human action recognition is one of the fundamental challenges in robotics systems. In this paper, we propose lightweight architecture based on deep neural networks just using RGB data. The proposed consists convolution network (CNN), long short-term memory (LSTM) units, and temporal-wise attention model. First, CNN used to extract spatial features distinguish objects from background with both local semantic characteristics. Second, two kinds LSTM are performed feature maps different layers...

The analysis of marine environmental parameters plays an important role in areas such as sea surface simulation modeling, clutter characteristics, and monitoring. However, ocean observation remote sensing satellites typically deliver large volumes data with limited spatial resolution, which often does not meet the precision requirements practical applications. To overcome challenges constructing high-resolution parameters, this study conducts a systematic comparison various interpolation...

Demosaicking is a digital image process to reconstruct full color images from incomplete samples an sensor. It unavoidable for many devices incorporating camera sensor (e.g., mobile phones, tablet, and so on). In this paper, we introduce new demosaicking algorithm based on polynomial interpolation-based demosaicking. Our method makes three contributions: calculation of error predictors, edge classification differences, refinement stage using weighted sum strategy. predictors are generated...

In this paper, an anisotropic holographic metasurface design is proposed, fabricated, and measured, to demonstrate that it can generate multiple beams with different orbital angular momentum (OAM) modes in the radio-frequency domain. The composed of array quasi-periodic square particles a rectangular slot upper metallic layer covered dielectric ground. classic leaky-wave theory microwave holography method are introduced construct holograms interfere surface waves excited by monopole antenna...

Human mesh recovery from one single image has achieved rapid progress recently, but many methods suffer the appearance overfitting since training data are collected along with accurate 3D annotations in controlled settings of monotonous backgrounds or simple clothes. Some regress human vertices poses to tackle above problem. However topologies have not been well exploited, and artifacts often generated. In this paper, we aim find an efficient low-cost solution reconstruction. To end, propose...

Wavelength-tunable continuous-wave and Q-switching visible Ho³⁺-doped all-fiber lasers emitting at green deep-red wavelengths are successfully demonstrated.

Traditional paper certificates and electronic have difficulties in preservation management, not to mention other problems concerning inconvenient verification, poor reliability, anti-counterfeiting anti-tampering. This proposes a scheme designed build decentralized certificate system that is based on blockchain technology smart contract, which set of aiming at providing services for college students' innovation entrepreneurship competition developed. In this system, certain functions the...

In this paper, a dual-band dual-polarized reflectarray for generating dual beams with respect to carrying two different orbital angular momentum (OAM) topological charges operating in the C-band horizontal polarization and X-band vertical is proposed, separate horns performing on proposed bands as feeding. The consists of band reflective element cells orthogonal directions. Owing composing elements orthogonally interleaved surface, corresponding phase compensation performance one can be...

Visual relationship detection (VRD) is one newly developed computer vision task, aiming to recognize relations or interactions between objects in an image. It a further learning task after object recognition, and important for fully understanding images even the visual world. has numerous applications, such as image retrieval, machine robotics, question answer (VQA), reasoning. However, this problem difficult since relationships are not definite, number of possible much larger than objects....

Recovery algorithms play a key role in compressive sampling (CS). Currently, popular recovery algorithm for CS is the orthogonal matching pursuit (OMP), which possesses merits of low complexity and good quality. Considering that OMP involves massive matrix/vector operations, it very suited to being implemented parallel on graphics processing unit (GPU). In this paper, we first analyze each module point out bottlenecks lie projection least-squares module. To speedup module, Fujimoto's...

This paper presents an unsupervised change detection approach for synthetic aperture radar images based on a fuzzy active contour model and genetic algorithm. The aim is to partition the difference image which generated from multitemporal satellite into changed unchanged regions. Fuzzy technique appropriate analyze where regions are not always statistically homogeneous. Since interval type-2 sets well-suited modeling various uncertainties in comparison traditional sets, they combined with...

In this Letter, to the best of our knowledge, a novel anisotropic surface impedance holographic metasurface antenna for generating orbital angular momentum (OAM) is developed through design, fabrication, and measurement at radio frequencies. The classical leaky-wave theory microwave holography method are combined realize vortex waves carrying different OAM modes flexibly. composed subwavelength quasi-periodic metallic patches on grounded dielectric substrate operates by exciting interference...

The study of sea clutter reflectivity plays an important role in radar performance evaluations the military industry. industrial bodies are trying to apply intelligent processing technology system with form Internet Things and Industry 4.0. Many models that have been proposed difficult fully adapt surrounding seas different systems China. This article proposes a model named multi-source input neural network (MSINN) using collected by ultra high frequency (UHF) radar. In order prepare data...

The evaporation duct is a weather phenomenon that often occurs in marine environments and affects the operation of shipborne radar. most important evaluation parameter height (EDH). Forecasting EDH adjusting working parameters modes radar system advance can greatly improve performance. Traditionally, short-term forecast methods have been used to estimate EDH, which are characterized by low time resolution poor accuracy. In this study, novel approach for nowcasting proposed based on deep...

The measurement and analysis of the interaction between Bessel vortex electromagnetic (EM) several standard targets are presented in this paper. With aid angular spectrum expansion (ASE) method physics optics (PO) theorem, scattering results on plates (metal dielectric) a sphere could be derived. Furthermore, plane near-field scanning near-far field conversion methods were implemented to compare theoretical radar cross section (RCS). In experiment, quasi wave was generated by holographic...

ABSTRACT Driven by the rise of internet, recent years have witnessed gradual manifestation commercial values online reviews. In movie industry, sentiment analysis serves as foundation for mining user preferences among diverse and multi‐layered audiences, providing insight into market value movies. As a representative task, aspect‐based (ABSA) aims to analyse extract fine‐grained elements their relations in terms discussed aspects. Relevant studies, particularly realm deep learning research,...

The infrared (IR) signature of an aerial target due to the reflection radiation from Sun, Earth's surface and atmosphere plays important role in detection tracking. As background surface, is distributed entire space a wide spectrum, it time-consuming obtain target's reflected radiation. This problem suitable for parallel implementation run on multicore CPU or many-core GPU because incident different directions each spectral wavelength can be calculated parallel. We consider three approaches:...

A low temperature supercritical fluid nitridation (SCF-nitridation) technique was investigated to dope nitrogen into a indium-tin-oxide (ITO) electrode boost the performance of hafnium oxide resistive random access memory (RRAM).

With the development of remote sensing technology, spatial and spectral resolutions hyperspectral images have become increasingly dense. In order to overcome difficulties in storage, transmission, manipulation images, an effective compression algorithm is requisite. The clustered differential pulse code modulation (C-DPCM), which a prediction-based image lossless algorithm, can achieve relatively high ratio, but its efficiency still requires improvement. This paper presents parallel...