Relational attention-based Markov logic network for visual navigation

The Journal of Supercomputing - We argue the agent’s low generalization problem for searching target object in challenging visual navigation could be solved by "how" and...     

Syntax–Aware graph convolutional network for the recognition of chinese implicit inter-sentence relations

The Journal of Supercomputing - In the literature, most previous studies on English implicit inter-sentence relation recognition only focused on semantic interactions, which could not exploit the...     

Joint Classification and Regression for Visual Tracking with Fully Convolutional Siamese Networks

International Journal of Computer Vision - Visual tracking of generic objects is one of the fundamental but challenging problems in computer vision. Here, we propose a novel fully convolutional...     

A Survey on Long-Tailed Visual Recognition

The heavy reliance on data is one of the major reasons that currently limit the development of deep learning. Data quality directly dominates the effect of deep learning models, and the long-tailed distribution is one of the factors affecting data quality. The long-tailed phenomenon is prevalent due to the prevalence of power law in nature. In this case, the performance of deep learning models is often dominated by the head classes while the learning of the tail classes is severely underdeveloped. In order to learn adequately for all classes, many researchers have studied and preliminarily addressed the long-tailed problem. In this survey, we focus on the problems caused by long-tailed data distribution, sort out the representative long-tailed visual recognition datasets and summarize some mainstream long-tailed studies. Specifically, we summarize these studies into ten categories from the perspective of representation learning, and outline the highlights and limitations of each category. Besides, we have studied four quantitative metrics for evaluating the imbalance, and suggest using the Gini coefficient to evaluate the long-tailedness of a dataset. Based on the Gini coefficient, we quantitatively study 20 widely-used and large-scale visual datasets proposed in the last decade, and find that the long-tailed phenomenon is widespread and has not been fully studied. Finally, we provide several future directions for the development of long-tailed learning to provide more ideas for readers.

     

Relation network based on multi-granular hypergraphs for person re-identification

Applied Intelligence - In recent years, person re-identification (re-ID) has become a widespread research topic that focuses on retrieving target pedestrians from a set of images, typically taken...     

Heterogeneous feature ensemble modeling with stochastic configuration networks for predicting furnace temperature of a municipal solid waste incineration process

Neural Computing and Applications - Considering the accuracy, generalization ability, stability, and training efficiency of a furnace temperature model in the process of municipal solid waste...     

Comparison of gas sensing properties based on hollow and porous α-Fe<Subscript>2</Subscript>O<Subscript>3</Subscript> nanotubes

Hollow and porous α-Fe₂O₃ nanotubes were successfully synthesized by single nozzle electrospinning method followed by annealing treatment. The crystal structures and morphologies of the as-prepared materials were characterized by X-ray diffraction and scanning electron microscopy, respectively. The as-prepared materials were applied to construct gas sensor devices which gas sensing properties were further investigated. The obtained results revealed that porous α-Fe₂O₃ nanotube gas sensors exhibit a markedly enhanced gas sensing performance compared with hollow α-Fe₂O₃ nanotube gas sensors, which was about three times higher to 100 ppm acetone at 240 °C. Interestingly, hollow and porous α-Fe₂O₃ nanotube gas sensors both showed fast response–recovery time and good selectivity, but the porous ones possessed the shorter recovery time. The improved properties could be attributed to the unique morphology of porous nanotubes. Thus, further improvement of performance in metal-oxide-semiconductors materials could be realized by preparation the unique porous structures of nanotubes. Moreover, it is expected that porous metal-oxide-semiconductors nanotubes could be further design as promising candidates for gas sensing materials.     

The interlaminar shear failure characteristics of asphalt pavement coupled heating cables

Nowadays, heating cables are used as heat sources for heating pavements in practical engineering. However, there is a contradiction between the snow melting function and the interlaminar stability of heating pavement. In order to solve the contradiction, the interlaminar failure behavior of asphalt mixture coupled heating cables specimen (AMCS) was researched, through experiments and the finite element method. Under the different conditions of heating cables and rolling times, a series of direct shear tests was performed at the interface of AMCS, to compare the interlaminar stability of three different AMCS. Meanwhile, based on the bilinear cohesive zone model and coulomb friction model a 2D finite element model was established, to simulate this shear failure processes and make up for the limitations of the experiment. According to above test and simulation results, the failure mechanism and the weakest interface in AMCS were found, and the influence of the heating cable’s diameter and embedded spacing on the interlaminar shear strength were found. Then, a modified coulomb theorem model was proposed to predict the shear strength of the AMCS. This research enriches the design theory of the heating pavement and it has great significance for its structural design of heating asphalt pavement.     

Monitoring the Dynamic Process of Formation of Plasmonic Molecular Junctions during Single Nanoparticle Collisions

The capability to study the dynamic formation of plasmonic molecular junction is of fundamental importance, and it will provide new insights into molecular electronics/plasmonics, single‐entity electrochemistry, and nanooptoelectronics. Here, a facile method to form plasmonic molecular junctions is reported by utilizing single gold nanoparticle (NP) collision events at a highly curved gold nanoelectrode modified with a self‐assembled monolayer. By using time‐resolved electrochemical current measurement and surface‐enhanced Raman scattering spectroscopy, the current changes and the evolution of interfacial chemical bonding are successfully observed in the newly formed molecular tunnel junctions during and after the gold NP “hit‐n‐stay” and “hit‐n‐run” collision events. The results lead to an in‐depth understanding of the single NP motion and the associated molecular level changes during the formation of the plasmonic molecular junctions in a single NP collision event. This method also provides a new platform to study molecular changes at the single molecule level during electron transport in a dynamic molecular tunnel junction.