MSF-Net: A Multilevel Spatiotemporal Feature Fusion Network Combines Attention for Action Recognition

An action recognition network that combines multi-level spatiotemporal feature fusion with an attention mechanism is proposed as a solution to the issues of single spatiotemporal feature scale extraction,information redundancy,and insufficient extraction of frequency domain information in channels in 3D convolutional neural networks.Firstly,based on 3D CNN,this paper designs a new multilevel spatiotemporal feature fusion(MSF)structure,which is embedded in the network model,mainly through multilevel spatiotemporal feature separation,splicing and fusion,to achieve the fusion of spatial perceptual fields and short-medium-long time series information at different scales with reduced network parameters;In the second step,a multi-frequency channel and spatiotemporal attention module(FSAM)is introduced to assign different frequency features and spatiotemporal features in the channels are assigned corresponding weights to reduce the information redundancy of the feature maps.Finally,we embed the proposed method into the R3D model,which replaced the 2D convolutional filters in the 2D Resnet with 3D convolutional filters and conduct extensive experimental validation on the small and medium-sized dataset UCF101 and the largesized dataset Kinetics-400.The findings revealed that our model increased the recognition accuracy on both datasets.Results on the UCF101 dataset,in particular,demonstrate that our model outperforms R3D in terms of a maximum recognition accuracy improvement of 7.2%while using 34.2%fewer parameters.The MSF and FSAM are migrated to another traditional 3D action recognition model named C3D for application testing.The test results based on UCF101 show that the recognition accuracy is improved by 8.9%,proving the strong generalization ability and universality of the method in this paper.

Simplified Compact Perovskite Solar Cells with Efficiency of 19.6% via Interface Engineering

For the commercialization of perovskite solar cells(PSCs), it is more appealing to develop high-performance simplified PSCs where perovskite films are just sandwiched between the back and front electrodes, in order to simplify the fabrication process and to reduce the cost. However, to date, this kind of devices shows rather low performance, and there are few researches on this subject.Herein, we report on a kind of compact PSCs(CPSCs) that are free of independent charge transport layers(CTLs). The devices are realized by the use of organic monolayer-modified effective electrodes, along with the use of [6,6]-phenyl-C61-butyric acid methyl ester(PCBM)-assisted anti-solvent technique to obtain ultra-thin(~10 nm) PCBM-embedded perovskite films. Compared to control devices, CPSCs achieve a promising champion power conversion efficiency of 19.6% with largely reduced hysteresis. Moreover, the unencapsulated CPSC shows good stability under ambient atmosphere, with only 10% efficiency loss after 60 days’ storage. This work indicates that, by delicate design, CPSCs with smaller materials consumption in device architecture can perform competitively as conventional PSCs. Further reduction in the actual usage of costly CTL materials can be expected upon our CPSCs by developing more facile and economic methods to prepare ultra-thin CTLs.

Multi-Label Feature Selection Based on Improved Ant Colony Optimization Algorithm with Dynamic Redundancy and Label Dependence

The world produces vast quantities of high-dimensional multi-semantic data.However,extracting valuable information from such a large amount of high-dimensional and multi-label data is undoubtedly arduous and challenging.Feature selection aims to mitigate the adverse impacts of high dimensionality in multi-label data by eliminating redundant and irrelevant features.The ant colony optimization algorithm has demonstrated encouraging outcomes in multi-label feature selection,because of its simplicity,efficiency,and similarity to reinforcement learning.Nevertheless,existing methods do not consider crucial correlation information,such as dynamic redundancy and label correlation.To tackle these concerns,the paper proposes a multi-label feature selection technique based on ant colony optimization algorithm(MFACO),focusing on dynamic redundancy and label correlation.Initially,the dynamic redundancy is assessed between the selected feature subset and potential features.Meanwhile,the ant colony optimization algorithm extracts label correlation from the label set,which is then combined into the heuristic factor as label weights.Experimental results demonstrate that our proposed strategies can effectively enhance the optimal search ability of ant colony,outperforming the other algorithms involved in the paper.

Enhanced oxidation and graphitization resistance of polycrystalline diamond sintered with Ti-coated diamond powders

To improve the oxidation and graphitization resistances of the polycrystalline diamond(PCD), Ti coating was deposited on the diamond powders via magnetic sputtering method, which achieved a uniform Ti C protection barrier in PCD during the sintering process. The phase compositions, microstructures and thermal stability of Ti-PCD were characterized by X-ray diffraction(XRD), Auger electron spectroscopy(AES),scanning electron microscopy(SEM) and thermal gravimetric-differential scanning calorimetry(TG-DSC).The results demonstrate that the oxidation and graphitization resistances of PCD are strengthened due to the existence of Ti C phase, which acts as an effective inhibitor. The as-received inhibitor delays the oxidation and graphitization of PCD, elevating their initial temperature by ~50°C and ~100°C, respectively. During the annealing treatment of Ti-PCD, the priory oxidation of Ti C, which produces Ti O2 as an oxygen barrier, postpones the diamond oxide. Moreover, the Ti C barrier also protects diamond grains from direct contact with cobalt, thus a lower cobalt-catalytic graphitization, and yields to an improved graphitization resistance of PCD. The enhanced oxidation and graphitization resistances of PCD are of significant importance for practical applications to elevated temperatures.

Sub-10-fs pulse generation directly from a KLM Ti: sapphire laser

By reducing the effects of frequency dependent mode size and intracavity gain saturation of a KLM Ti: sapphire laser, the pulse bandwidth broadening is enhanced significantly. 8.5 fs pulses with the bandwidth of 92 nm at wavelength of 710 nm have been generated directly from this laser and measured with an extracavity group velocity dispersion compensation system.

Polyoxometalate-Induced Efficient Recycling of Waste Polyester Plastics into Metal–Organic Frameworks

Polyester plastics such as poly(ethylene terephthal-ate)(PET)are utilized commonly in everyday life,yet only a small portion of these plastics are recycled,and typically,the recycling procedures face energy or pollution problems.

A maximum flow algorithm for buffer-limited delay tolerant networks

Deep space networks,satellite networks,ad hoc networks,and the Internet can be modeled as DTNs(Delay Tolerant Networks).As a fundamental problem,the maximum flow problem is of vital importance for routing and service scheduling in networks.However,there exists no permanent end-to-end path since the topology and the characteristics of links are time-variant,resulting in a crucial maximum flow problem in DTNs.In this paper,we focus on the single-source-single-sink maximum flow problem of buffer-limited DTNs,followed by a valid algorithm to solve it.First,the BTAG(Buffer-limited Time Aggregated Graph)is constructed for modeling the buffer-limited DTN.Then,on the basis of BTAG,the two-way cache transfer series and the relevant transfer rules are designed,and thus a BTAG-based maximum flow algorithm is proposed to solve the maximum flow problem in buffer-limited DTNs.Finally,a numerical example is given to demonstrate the effectiveness of the proposed algorithm.