As an important part of intelligent surveillance systems, person re-identification (PReID) has drawn wide attention the public in recent years. Many deep learning-based PReID methods have used or multi-scale feature learning modules to enhance discrimination learned features. However, mechanisms may lose some information. Moreover, models usually embed module into backbone network, which increases complexity testing network. To address two issues, we propose a supervision with model for...

Image classification is a very important tool for remotely sensed hyperspectral image processing. Techniques able to exploit the rich spectral information contained in data, as well its spatial-contextual information, have shown success recent years. Due high dimensionality of spectral-spatial techniques are quite demanding from computational viewpoint. In this paper, we present computationally efficient parallel implementation method based on spatially adaptive Markov random fields (MRFs)....

Heterostructured photocatalysts play a significant role in the removal of contaminants by decreasing recombination photo-induced charges. Herein, we presented novel TiO2/C/BiVO4 ternary hybrids employing 2D layered Ti3C2 MXene precursor, overcoming lattice mismatching TiO2/BiVO4 binary heterostructures simultaneously. Raman and XPS analyses proved strong coupling effects TiO2, carbon BiVO4 components, were identified from high-resolution transmission electron microscopy results. Moreover,...

The sky region in an image provides horizontal and background information for autonomous ground robots is important vision-based robot navigation. This paper proposes a detection algorithm within single based on gradient energy function optimization. Unlike most existing methods, the proposed applicable to both colour greyscale images. Firstly, of obtained. Then, optimal segmentation threshold domain calculated according optimization preliminary estimated. Finally, post-processing method...

In this letter, a real-time implementation of the logistic regression via variable splitting and augmented Lagrangian (LORSAL) algorithm for sparse multinomial is presented on commodity graphics processing units (GPUs) using Nvidia's compute unified device architecture. The proposed parallel method properly exploits GPU architecture at low level, including its shared memory, takes full advantage computational power GPUs to achieve classification performance hyperspectral images first time in...

This paper proposes a dual-stream 3D space-time convolutional neural network action recognition framework. The original depth map sequence data is set as the input in order to study global characteristics of each category. high correlation within human itself considered time domain, and then deep motion introduced another stream network. Furthermore, corresponding skeleton third whole Although has advantage including information, it also confronted with problems existence rate change,...

Recently, in addition to petrophysical properties, the present-day situ stress (PDIS) is also considered as one of controlling factors for highly heterogeneous deep and ultradeep carbonate reservoirs. However, specific relationship between PDIS distribution reservoirs remains unclear. This situation restricts our understanding law accurate prediction favorable areas. paper focuses on unraveling orientation, magnitude, its influence Upper Member Yingshan Formation within S region Tahe...

Spatial-spectral classification is a very important topic in the field of remotely sensed hyperspectral imaging. In this work, we develop parallel implementation novel supervised spectral-spatial classifier, which models likelihood probability via l <sub xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">1</sub> - xmlns:xlink="http://www.w3.org/1999/xlink">2</sub> sparse representation and spatial prior as Gibbs distribution. This classifier takes...

Classification is one of the most important analysis techniques for hyperspectral image analysis. Sparse representation an extremely powerful tool this purpose, but high computational complexity sparse representation-based classification limits their application in time-critical scenarios. To improve efficiency and performance analysis, paper develops a new parallel implementation on graphics processing units (GPUs). First, optimized model based spatial correlation regularization spectral...

Self-supervised representation learning for videos has been very attractive recently because these methods exploit the information inherently obtained from video itself instead of annotated labels that is quite time-consuming. However, existing ignore importance global observation while performing spatio-temporal transformation perception, which highly limits expression capabilities representation. This paper proposes a novel pretext task combines temporal perception with motion amplitude...

Melatonin plays a critical role in the pathophysiological process including circadian rhythm, apoptosis, and oxidative stress. It can be synthesized ocular tissues, its receptors are also found eye, triggering more investigations concentrated on of melatonin eye. In past decades, protective therapeutic potentials for diseases have been widely revealed animal models. Herein, we construct knowledge map treating through bibliometric analysis review current understanding clinical evidence. The...

Beryllium (Be) has been selected as the solid neutron multiplier material for a tritium breeding blanket module in ITER, which is also primary option of Chinese TBM program. But irradiation swelling beryllium severe under high temperature, damage and doses transmutation-induced helium. Advanced multipliers with stability at temperature are desired demonstration power plant (DEMO) reactors China Fusion Engineering Test Reactor (CFETR). alloys mainly composed Be12M (M W or Ti) phase were...

The classical person re-identification methods are mostly focused on employing discriminative features amongst which the distance is measured Euclidean space, while effort of re-ranking constrained as lack utilization quality context representation in embedding set.In this paper, we incorporate graph models feature subsets resorting to initial ranking by adopting integration attention mechanism into convolution network.On one hand, information regarding pairs considered compute group...

For 3D action recognition, the main challenge is to extract long-range semantic information in both temporal and spatial dimensions. In this paper, order better excavate from large number of unlabelled skeleton sequences, we propose Self-supervised Spatial-temporal Representation Learning (SSRL), a contrastive learning framework learn representation. SSRL consists two novel inference tasks that enable network global dimensions, respectively. The task learns persistence human actions through...

Short-term hourly reliable prediction of significant wave height is an important research topic in coastal engineering. Many researchers have carried out in-depth studies many ocean regions. Generally, most this work implemented through numerical models. However, as for models, with the increase duration, accumulation randomness leads to poor effect. In paper, four buoy stations Taiwan Strait are taken objects. We propose a algorithm, which combines weather model WRF and deep-learning model,...

Semi-supervised learning is a pattern that can utilize labeled data and unlabeled to train deep neural networks. In semi-supervised methods, self-training-based methods do not depend on augmentation strategy have better generalization ability. However, their performance limited by the accuracy of predicted pseudo-labels. this paper, we propose reduce noise in pseudo-labels from two aspects: predictions confidence predictions. For first aspect, similarity graph structure (SGSL) model...

Behavior sequences are generated by a series of spatio-temporal interactions and have high-dimensional nonlinear manifold structure. Therefore, it is difficult to learn 3D behavior representations without relying on supervised signals. To this end, self-supervised learning methods can be used explore the rich information contained in data itself. Context-context contrastive construct embedded Euclidean space distance relationship between data, find geometric distribution data. However,...