Emotion recognition from facial expressions is an interesting and challenging problem has attracted much attention in recent years. Substantial previous research only been able to address the ambiguity of “what describes expression”, which assumes that each expression associated with one or more predefined affective labels while ignoring fact multiple emotions always have different intensities a single picture. Therefore, depict accurately, this paper adopts label distribution learning...

Microplastics (MPs) have been found in nearly half of the fish samples collected from field worldwide. There are feeding behavior differences caused by animal personality among individuals. However, how affects exposure to microplastics is unclear. In this study, adult zebrafish were used as a model study effects (boldness) on MPs exposure. Experiment showed that captured actively when added system but spit part out after capturing. Bold had higher activities, more frequently and swallowed...

Among various elements of manga, character's face plays one the most important role in access and retrieval. We propose a DNN-based method to do manga detection, which is challenging but relatively unexplored topic. Given page, we first find candidate regions based on selective search scheme. A deep neural network then proposed detect faces appearance. evaluate large-scale benchmark, show performance comparison convincing evaluation results that have rarely done before.

Fresh water cyanobacterial algal blooms represent a major health risk because these organisms produce cylindrospermopsin, toxic, structurally complex, zwitterionic uracil-guanidine alkaloid recognized by the EPA as dangerous drinking contaminant. At present, ability to detect and quantify presence of cylindrospermospin in samples is severely hampered lack an isotopically labeled standard for analytical mass spectrometry. Herein, we present concise, scaled total synthesis 15N...

Denitrifying anaerobic methane oxidation (damo) bioprocesses can remove nitrate using as the electron donor, which gains great concern due to current stringent discharge standard of nitrogen in wastewater treatment plants. To obtain an engineering acceptable removal rate (NRR) and demonstrate long-term stable ability damo system under conditions methane, two sequencing batch reactors (SBRs) fed with only were operated for more than 600 days at 30 °C. The NRR 21.91 ± 0.73 mg NO

Label enhancement (LE) is a procedure of recovering the label distributions from logical labels in multi-label data, purpose which to better represent and mine ambiguity problem through form distribution. Existing LE work mainly concentrates on how leverage topological information feature space correlation among labels, all are based single view data. In fact that there many multi-view data real-world applications, can provide richer semantic different perspectives, this paper first presents...

WiFi fingerprint-based indoor positioning system emerges to provide fundamental location-related service in heterogeneous communications networks. It relies on crowdsourcing technology the collection of received signal strength (RSS) dynamically update fingerprint database. However, this crowdsourced is vulnerable intrusion dishonest users (i.e., attackers). Attackers may manipulate number who submit RSS fingerprints, and finally mislead evaluation crowd traffic evaluation. In paper, we...

Underwater image enhancement recovers degraded underwater images to produce corresponding clear images. Image methods based on deep learning usually use paired data train the model, while such data, e.g., and images, are difficult capture simultaneously in environment. In addition, how retain detailed information well enhanced is another critical problem. To solve issues, we propose a novel unpaired method via cycle generative adversarial network (UW-CycleGAN) recover Our proposed...

In this paper, we propose a new method for detecting abnormal human behavior based on skeleton features using self-attention augment graph convolution. The data have been proved to be robust the complex background, illumination changes, and dynamic camera scenes are naturally constructed as in non-Euclidean space. Particularly, establishment of spatial temporal convolutional networks (ST-GCN) can effectively learn spatio-temporal relationships Non-Euclidean Structure Data. However, it only...

In this paper, we focuses on an access control issue in the Internet of Things (IoT). Generally, firstly propose a decentralized IoT system based blockchain. Then establish secure fine-grained strategies for users, devices, data, and implement with smart contract. To trigger contract, design different transactions. Finally, use multi-index table struct right's establishment, store right into Key-Value database to improve scalability system. addition, security also records blockchain database.

Intelligent video surveillance in modern city is becoming more and popular, higher quality of images required automatic analysis recognition. In this paper, we develop a kind haze-free method for outdoor foggy image with low contrast. The based on the multi-scale guided filtering combined Retinex algorithm. Compared traditional series Retinex, it may better following aspects: process details enhancement, color fidelity overcome noise etc. experimental results show advantages our algorithm...

Label distribution learning has attracted more and attention in view of its generalized ability to express the label ambiguity. However, it is much expensive obtain information data rather than logical labels. Thus, enhancement proposed recover distributions from In this paper, we propose a novel method by using privileged information. We first apply multi-label model implicitly capture complex structural between instances generate Second, adopt LUPI (learning with information) paradigm...

<title>Abstract</title> Computer haptics (CH) is about integration of tactile sensation and rendering in Metaverse. However, unlike computer vision (CV) where both hardware infrastructure software programs are well developed, a generic data capturing device that serves the same role as what camera does for CV, missing. Bioinspired by electrophysiological processes human somatosensory nervous system, here we propose scanner together with neuromorphically-engineered which closed-loop...

Despite the great progress in 3D pose estimation from videos, there is still a lack of effective means to extract spatio-temporal features different granularity complex dynamic skeleton sequences. To tackle this problem, we propose novel, skeleton-based U-Net(STUNet) scheme deal with multiple scales for human video. The proposed STUNet architecture consists cascade structure semantic graph convolution layers and structural temporal dilated layers, progressively extracting fusing fine-grained...

A novel molecular device (trans-azobenzene embedded N-(11-pyrenyl methyl)aza-21-crown-7) with double functional devices was designed on the basis of theoretical calculations. Pyrenyl methyl covalently bonded to aza-21-crown-7 at nitrogen position interacting a series alkaline-earth metal cations (Mg2+, Ca2+, Sr2+, and Ba2+) investigated. The fully optimized geometries real frequency calculations were investigated using computational strategy based density theory B3LYP/6-31G(d) level. Free...

Visual comfort is an important component for stereoscopic image quality and viewing experience. In recent years, it has been widely investigated through feature analysis different kinds of subjective experiments. this paper, according to the existing works objective evaluation model, we have screened extracted thirteen disparity-based content-based features image. To find relationship between these given mean opinion score(MOS) each image, method Random Forest (RF) Regressor was used...