Search

This study was designed to investigate the associations among motives and physical activity behaviors and health in urban college students in China. Eight hundred and eighty-seven college students (521 females; M[subscript age] = 20.51, SD = ± 1.67) were recruited from four universities in South and South-central China. Participants' motives (i.e., interest/enjoyment, competence, appearance, fitness, social) toward to physical activity behaviors were measured by the established Motives for Physical Activities Measure (Ryan et al., 1997), while physical activity behaviors were assessed via the validated International Physical Activity Questionnaires for Chinese (Macfarlane et al., 2007). The data were collected in June 2016 and 2017. Participants' total metabolic equivalent (METs) was calculated as the outcome of physical activity behaviors. On average, participants reported 412 METs (SD = 290.82) of physical activity for the prior week. Correlation analysis suggested that all motive components were moderately or highly correlated with one another (r = 0.37 -0.74; p < 0.01), and these variables were significantly, yet modestly, related to physical activity behaviors (r = 0.12 -0.24; p < 0.05). Regression analyses further indicated that the whole model explained 24.5% of the variance [F (5, 394) = 5.02, p < 0.01] when using motives to predict physical activity behaviors. Interestingly, participants' interest/enjoyment was the only significant and positive predictor for their physical activity (ß= 0.23, p < 0.01). Findings suggest that Chinese college students were relatively physically active on a weekly basis. Students' interest/enjoy toward physical activity is a very important motive in promoting physical activity behavior among this population. Health professionals and educators are encouraged to offer a variety of interesting and enjoyable activities for Chinese college students with the goal of helping this population develop and maintain a physically active lifestyle.

Nighttime semantic segmentation is an important but challenging research problem for autonomous driving. The major challenges lie in the small objects or regions from the under-/over-exposed areas or suffer from motion blur caused by the camera deployed on moving vehicles. To resolve this, we propose a novel hard-class-aware module that bridges the main network for full-class segmentation and the hard-class network for segmenting aforementioned hard-class objects. In specific, it exploits the shared focus of hard-class objects from the dual-stream network, enabling the contextual information flow to guide the model to concentrate on the pixels that are hard to classify. In the end, the estimated hard-class segmentation results will be utilized to infer the final results via an adaptive probabilistic fusion refinement scheme. Moreover, to overcome over-smoothing and noise caused by extreme exposures, our model is modulated by a carefully crafted pretext task of constructing an exposure-aware semantic gradient map, which guides the model to faithfully perceive the structural and semantic information of hard-class objects while mitigating the negative impact of noises and uneven exposures. In experiments, we demonstrate that our unique network design leads to superior segmentation performance over existing methods, featuring the strong ability of perceiving hard-class objects under adverse conditions. [ABSTRACT FROM AUTHOR]

This study aims to enhance the high-temperature resistance of ultrahigh-performance concrete (UHPC) by optimizing the cement-fly ash-silica fume ternary cement matrix composition and the geometric characteristics of the plastic fibers. The UHPC was designed and prepared with the cement-fly ash-silica fume ternary cement matrix through simplex centroid method and 2% volume fraction steel fiber. Polypropylene fibers with different geometric characteristics were incorporated to enhance the high-temperature resistance. The fire resistance is examined through the high-temperature exposure test. The initial bursting temperature is recorded, and the deterioration of mechanical performance is characterized. It was found that the decrepitation temperature of UHPC first increases and then decreases with the fly ash (FA) content. Meanwhile, the influence of the silica fume (SF) and cement content on the decrepitation temperature is not obvious. The optimum mix ratio of 50%–65% cement, 20%–30% FA, and 10%–20% SF is recommended to prepared the cementitious matrix with 450°C or higher initial burst temperature. The crack mouth opening displacement (CMOD) test assisted with digital image correlation (DIC) examination is conducted to characterize the fracture performance before and after high-temperature exposure. It was found that the fracture resistance of UHPC first increases and then decreases with the silica fume content, which reaches the maximum value of 16.64 N/mm with 27% SF content. Meanwhile the high-temperature resistance of UHPC increases and then decreases with the increase of the length of the doped polypropylene (PP) fibers. The added 1.2% volume fraction PP fiber can resolve the spalling issues of the UHPC materials, and the residual compressive strengths after exposure to 1,000°C of UHPC samples containing 15 mm–18 μm and 15 mm–33 μm fibers can exceed 50 MPa. This study can serve as a solid base for the fire resistance design of UHPC materials in field construction. [ABSTRACT FROM AUTHOR]

By the finite element method, the structural vibration response is calculated under the action of the axial exciting force and the moment with different distribution form, and then the transfer function of the mean square normal velocity is analyzed. The wavenumber spectrum analysis method is used to separate and quantify the shell vibration in the wavenumber domain, and then the relation between the structural vibration characteristics and the structural wavelength is summarized. It is concluded that the structural vibration and radiated noise can be reduced under the symmetric action of axial exciting force and the moment. Based on the above conclusion, a symmetrical thrust bearing supporting system is designed and the stiffness of the supporting structure in the axial direction is controlled by selecting suitable size of structural members, therefore, the structural vibration and radiated noise of the submarine is reduced significantly.

Objectives To improve underwater acoustic communication and sonar detection ability in different seasons in an actual coastal environment, this study focuses on the optimum depth of shallow water acoustic propagation.Methods Using normal mode methods, incoherent transmission loss versus depth and range is calculated with different source locations and frequencies under typical sound speed profiles, and the effects of different sound speed profiles on the optimum depth are studied.Results Optimum depth at low frequencies does not change with sound speed profiles and source locations; it is always located near the middle of the water layer. At high frequencies, when the source is located at the isothermal layer of thermocline shallow water or a water layer with a constant gradient, the optimum depth is equal to the source depth and its conjugate depth within the iso-velocity layer. The optimum depth of positive and negative gradient shallow water is the same as the source depth.Conclusions The results of this study can provide useful references for the selection of receiver depth and passive sonar.

This paper introduces a method for high-resolution lattice image reconstruction and dislocation analysis based on diffraction extinction. The approach primarily involves locating extinction spots in the Fourier transform spectrum (reciprocal space) and constructing corresponding diffraction wave functions. By the coherent combination of diffraction and transmission waves, the lattice image of the extinction planes is reconstructed. This lattice image is then used for dislocation localization, enabling the observation and analysis of crystal planes that exhibit electron diffraction extinction effects and atomic jump arrangements during high-resolution transmission electron microscopy (HRTEM) characterization. Furthermore, due to the method's effectiveness in localizing dislocations, it offers a unique advantage when analyzing high-resolution images with relatively poor quality. The feasibility of this method is theoretically demonstrated in this paper. Additionally, the method was successfully applied to observed edge dislocations, such as 1 / 6 [ 21 1 − ] , 1 / 6 [ 2 − 1 1 − ] , and 1 / 2 [ 0 1 − 1 ] , which are not easily observable in conventional HRTEM characterization processes, in electro-deposited Cu thin films. The Burgers vectors were determined. Moreover, this paper also attempted to observe screw dislocations that are challenging to observe in high-resolution transmission electron microscopy. By shifting a pair of diffraction extinction spots and superimposing the reconstructed images before and after the shift, screw dislocations with a Burgers vector of 1 / 2 [ 01 1 − ] were successfully observed in electro-deposited Cu thin films. [ABSTRACT FROM AUTHOR]

[Objectives] In this paper, the influence of thrust bearing seating on the acoustic radiation of a submarine is studied. [Methods] By utilizing the Finite Element(FE)method and FE coupling fluid interface method, the structure vibration of the pressure hull is calculated with different distributions of longitudinal exciting forces. In this case, the symmetrical distribution of exciting forces and moments are beneficial for reducing vibration. According to the above conclusions, symmetrical thrust bearing seating is designed and the longitudinal stiffness of the system controlled through choosing the size of components. Two submarines are designed, one with conventional thrust bearing seating and the other one with symmetrical thrust bearing seating. The vibration response and radiation noise of the structures of the two submarines are forecasted. [Results] It can been seen that when the propeller longitudinal force is constant, symmetrical thrust bearing seating can significantly decrease the level of vibration and radiation noise. [Conclusions] The results can provide references for the acoustic optimization and design of the thrust bearing seating of submarines.

Statistical Energy Analysis(SEA)is an effective method for solving high frequency structural vibration and acoustic radiation problems. When we use it to analyze submerged structures, it is necessary to consider the actions of fluid as'heavy fluid' relative to structures, which differs from when it is used in the air. The simple model of a submerged cylindrical shell is used to calculate at a higher frequency using FEM/BEM. The SEA and FEM method are then used to calculate the radiation sound pressure level, verifying the accuracy of the SEA prediction for submerged structures. The classified method of subsystems and the effect of the error of the internal loss factor on the accuracy of the results are explored. The calculated results of SEA and FEM/BEM are very different below 400 Hz, and basically the same above 400 Hz. The error caused by the division of different subsystems is about 5 dB. The error in the calculation results caused by the error of the internal loss factor is 2-3 dB. It is possible to use SEA to calculate the radiated noise of an underwater cylindrical shell when the modal density is high enough.For the cylindrical shell, dividing the subsystems along the circumference is not reliable at a low frequency, as it may lead to inaccurate calculation results. At a high frequency, it is more accurate to divide the subsystems along the circumference than the axle. For subsystems with high energy, the internal loss factor has a greater effect on the simulation results, so a more accurate way should be taken to determine the internal loss factor of subsystems with high energy.

Due to the profitability and selfishness of crowdfunding system users, under fixed budget conditions, there are problems, such as low task completion rate due to insufficient participants and low data quality. However, the existing incentive mechanisms are mainly based on traditional economics, which believes that whether users participate in tasks depends on whether the benefits of the task outweigh the costs. Behavioral economics shows that people judge the value of gains and losses according to a reference point. The weight given to losses is more important than the weight given to the same gains. Therefore, this article considers the impact of reference dependency and loss aversion on user decision-making and proposes a participant selection mechanism based on reference dependency (PSM-RD) and a quality assurance mechanism based on loss aversion (QAM-LA). PSM-RD uses reference points to influence user pricing and selects more participants based on relative value. QAM-LA pays additional rewards based on the data quality of participants and motivates them to improve data quality by reconstructing utility functions. The simulation results show that compared with the ABSee mechanism, data quality has improved by 17%, and the value of completed tasks has increased by at least 40%. [ABSTRACT FROM AUTHOR]

An essential precondition for the effective use of low-frequency spread-spectrum acoustic signals is their synchronous acquisition. Due to the low bit rate that low-frequency spread-spectrum signals have, the length of the spreading spectrum code and the number of intra-chip carriers need to be precisely designed to balance the acquisition performance and the bit rate in low-frequency spread-spectrum signals. Furthermore, the selection of the acquisition method and sampling frequency depends on the specific application and system requirements, which will directly affect the processing speed and accuracy. Firstly, this study uses a cyclical stepping search combined with a fixed threshold and maximum correlation discriminant method to improve the FFT acquisition algorithm with a low Doppler frequency. Secondly, the effects of the spreading spectrum code parameters and sampling frequency on the acquisition performance are also investigated through simulation and experiments with acoustic sensors. The results show that both lengthening the spreading spectrum code and increasing the number of intra-chip carriers can greatly improve the acquisition performance. Increasing the sampling frequency can improve the ranging accuracy but has a very limited improvement effect on the acquisition performance. [ABSTRACT FROM AUTHOR]

Many studies have explored the efficacy of probiotics on autism spectrum disorder (ASD) in children, but there is no consensus on the curative effect. This systematic review and meta-analysis aimed to comprehensively investigate whether probiotics could improve behavioral symptoms in children with ASD. A systematic database search was conducted and a total of seven studies were included in the meta-analysis. We found a nonsignificant overall effect size of probiotics on behavioral symptoms in children with ASD (SMD = −0.24, 95% CI: −0.60 to 0.11, p = 0.18). However, a significant overall effect size was found in the subgroup of the probiotic blend (SMD = −0.42, 95% CI: −0.83 to −0.02, p = 0.04). Additionally, these studies provided limited evidence for the efficacy of probiotics due to their small sample sizes, a shorter intervention duration, different probiotics used, different scales used, and poor research quality. Thus, randomized, double-blind, and placebo-controlled studies following strict trial guidelines are needed to precisely demonstrate the therapeutic effects of probiotics on ASD in children. [ABSTRACT FROM AUTHOR]

Purpose: Current evidence regarding the use of commercially available immersive virtual reality (VR) headsets and compatible VR exercise apparatus in the promotion of individuals' physical activity (PA) is lacking. This exploratory study investigated the acute effects of a VirZoom VR exercise bike (vBike) on college students' physical responses compared to an exergaming bike (eBike) and a traditional stationary exercise bike (tBike). Method: Forty-eight college students (34 females, Xage = 23.58 ± 3.39; XBMI = 23.81 ± 3.57) completed three separate 20-minute cycling sessions on the vBike, eBike, and tBike in a counterbalanced order. Outcomes included rating of perceived exertion (RPE), assessed at 4-minute intervals throughout each cycling session, and overall PA, operationalized as pedal revolution counts (PRC) during each session. Results: Linear models indicated significant group differences between exercise modalities for RPE (F (2,138) = 14.76–27.30, all p <.001, η2 = 0.17–0.27) and PRC (F (2,138) = 74.862, p <.001, η2 = 0.48). Pairwise comparisons revealed that RPE was reported significantly higher at each time point (i.e., 4th minute, 8th minute, 12th minute, 16th minute, and 20th minute) during the tBike session vs. the vBike and eBike sessions (p <.001) and that participants had higher PRC during the vBike session vs. the eBike and tBike sessions (p <.001). Conclusion: A commercially available VR exercise bike would trigger greater overall PA as compared to exergaming and traditional cycling while being perceived less intense. Future intervention studies conducted in real-world settings to determine cause and effect relationship are warranted. [ABSTRACT FROM AUTHOR]

In this paper, we introduce a novel yet challenging research problem, interactive crowd video generation, committed to producing diverse and continuous crowd video, and relieving the difficulty of insufficient annotated real-world datasets in crowd analysis. Our goal is to recursively generate realistic future crowd video frames given few context frames, under the user-specified guidance, namely individual positions of the crowd. To this end, we propose a deep network architecture specifically designed for crowd video generation that is composed of two complementary modules, each of which combats the problems of crowd dynamic synthesis and appearance preservation respectively. Particularly, a spatio-temporal transfer module is proposed to infer the crowd position and structure from guidance and temporal information, and a point-aware flow prediction module is presented to preserve appearance consistency by flow-based warping. Then, the outputs of the two modules are integrated by a self-selective fusion unit to produce an identity-preserved and continuous video. Unlike previous works, we generate continuous crowd behaviors beyond identity annotations or matching. Extensive experiments show that our method is effective for crowd video generation. More importantly, we demonstrate the generated video can produce diverse crowd behaviors and be used for augmenting different crowd analysis tasks, i.e., crowd counting, anomaly detection, crowd video prediction. Code is available at https://github.com/Icep2020/CrowdGAN. [ABSTRACT FROM AUTHOR]

Data representation learning is one of the most important problems in machine learning. Unsupervised representation learning becomes meritorious as it has no necessity of label information with observed data. Due to the highly time-consuming learning of deep-learning models, there are many machine-learning models directly adapting well-trained deep models that are obtained in a supervised and end-to-end manner as feature abstractors to distinct problems. However, it is obvious that different machine-learning tasks require disparate representation of original input data. Taking human action recognition as an example, it is well known that human actions in a video sequence are 3-D signals containing both visual appearance and motion dynamics of humans and objects. Therefore, the data representation approaches with the capabilities to capture both spatial and temporal correlations in videos are meaningful. Most of the existing human motion recognition models build classifiers based on deep-learning structures such as deep convolutional networks. These models require a large quantity of training videos with annotations. Meanwhile, these supervised models cannot recognize samples from the distinct dataset without retraining. In this article, we propose a new 3-D deconvolutional network (3DDN) for representation learning of high-dimensional video data, in which the high-level features are obtained through the optimization approach. The proposed 3DDN decomposes the video frames into spatiotemporal features under a sparse constraint in an unsupervised way. In addition, it also can be regarded as a building block to develop deep architectures by stacking. The high-level representation of input sequential data can be used in multiple downstream machine-learning tasks, we evaluate the proposed 3DDN and its deep models in human action recognition. The experimental results from three datasets: 1) KTH data; 2) HMDB-51; and 3) UCF-101, demonstrate that the proposed 3DDN is an alternative approach to feedforward convolutional neural networks (CNNs), that attains comparable results. [ABSTRACT FROM AUTHOR]

Purpose. The bidirectional associations among children's motor skill competence (MSC), perceived competence (PC), physical activity (PA), and cardiorespiratory fitness (CRF) over time remain unanswered. This study is aimed at discerning the bidirectional relationships among elementary school children's MSC, PC, PA and, CRF over the course of one school year. Methods. A total of 261 second and third grade children (127 boys, 134 girls; mea n age = 8.27 years; BMI = 18.22 ± 3.71) were recruited from two Texas elementary schools. Approximately 73.56% of participants were White American. Children's baseline data were assessed in September/October in 2012 (Time1), and identical assessments were conducted in April/May in 2013 (Time2). MSC was assessed using product-oriented skill tests (e.g., throw, kick, and jump). PC was assessed via the Pictorial Scale of Perceived Competence for Children. Minutes spent in moderate-to-vigorous PA (MVPA) was assessed using ActiGraph GT3X+ accelerometers for five days, and CRF was assessed by the PACER test. Six age- and body mass index-adjusted cross-lagged panel models were used to test the relationships between the variables. Results. We observed that T1 MSC significantly predicted T2 MSC (β = 0.59 ; p < 0.01), T2 CRF (β = 0.28 ; p < 0.01), and T2 MVPA (β = 0.18 ; p < 0.01). Children's CRF was a positive predictor for T2 CRF (β = 0.56 ; p < 0.01) and T2 MSC (β = 0.13 ; p < 0.05) Additionally, T1 MVPA significantly predicted T2 MVPA (β = 0.30 ; p < 0.01) and T2 PC (β = − 0.14 ; p < 0.05). Conclusion. Findings suggested a fully bidirectional relationship between elementary children's MSC and CRF. Other bidirectional relationships among the variables were only partially supported. Educators and health professionals need to emphasize the importance of developing both MSC and CRF to maintain physical health over time. [ABSTRACT FROM AUTHOR]

Objective Purpose. The following review synthesized the currently available literature evaluating the effectiveness of smartphone-based physical activity applications on various health outcomes including physical and physiological outcomes, weight-related outcomes, and psychosocial outcomes. Design. A systematic review. Data Sources. 114 studies were gathered using the following search descriptors: ("mobile phone" OR "smartphone" OR "cell phone" OR "mobile device" OR "mobile apps" OR "mHealth") AND ("exercise" OR "physical activity" OR "physical fitness" OR "motor activity") AND ("physiological outcomes" OR "weight outcomes" OR "psychological outcomes" OR "health" OR "health behavior"). Seven databases were used including databases such as Academic Search Premier and PubMed. PRISMA guidelines were followed in this review. Eligibility Criteria for Selecting Studies. The 20 articles included in this review met the following inclusion criteria: (1) randomized and controlled trials, (2) involving an outcome variable measured by accelerometer, and (3) intervention enforced by a smartphone application. Results. Overall, 56% of the studies reviewed in this paper resulted in successful interventions. Of the 19 articles that examined the first individual health outcome of physical and physiological outcomes, 11 interventions resulted in a positive effect on one of the following parameters: MVPA/step count, sedentary behavior, cardiorespiratory fitness, and blood pressure. Six interventions examined the effects on the second individual health outcome, weight-related outcomes. Five of these interventions observed significant positive effects from mobile application interventions on weight and waist circumference. Six articles evaluated the effectiveness of smartphone-based physical activity interventions on the third and final individual health outcome, psychological outcomes, with four resulting in significant positive outcomes in self-efficacy, life enjoyment/satisfaction, and intrinsic PA motivation. Conclusion. The findings in this review suggest that mobile application physical activity interventions, compared to unguided exercise activities, can effectively improve certain health outcomes for individuals such as physical/physiological and weight-related outcomes. It was found that research in the area of effectiveness of mobile application interventions on specific psychosocial health outcomes such as self-efficacy, life enjoyment, and intrinsic PA motivation is limited. Thus, the effect of mobile health applications remains unclear for psychosocial outcomes. Due to this limitation, more research is warranted to confirm the findings of this review. [ABSTRACT FROM AUTHOR]

Synthesizing high dynamic range (HDR) images from multiple low-dynamic range (LDR) exposures in dynamic scenes is challenging. There are two major problems caused by the large motions of foreground objects. One is the severe misalignment among the LDR images. The other is the missing content due to the over-/under-saturated regions caused by the moving objects, which may not be easily compensated for by the multiple LDR exposures. Thus, it requires the HDR generation model to be able to properly fuse the LDR images and restore the missing details without introducing artifacts. To address these two problems, we propose in this paper a novel GAN-based model, HDR-GAN, for synthesizing HDR images from multi-exposed LDR images. To our best knowledge, this work is the first GAN-based approach for fusing multi-exposed LDR images for HDR reconstruction. By incorporating adversarial learning, our method is able to produce faithful information in the regions with missing content. In addition, we also propose a novel generator network, with a reference-based residual merging block for aligning large object motions in the feature domain, and a deep HDR supervision scheme for eliminating artifacts of the reconstructed HDR images. Experimental results demonstrate that our model achieves state-of-the-art reconstruction performance over the prior HDR methods on diverse scenes. [ABSTRACT FROM AUTHOR]

Objective. This systematic review synthesized current randomized controlled trials (RCTs) examining casual evidence regarding the effects of traditional and exergaming-based physical activity (PA) interventions on motor skill development in typically developed children (i.e., those aged 6-12 years). Methods. We adhered to the PRISMA-P statement and searched electronic databases (Medline, PsycInfo, Web of Science, PubMed, ERIC, Scopus, and SportDiscus) from inception through July 2020. We screened for peer reviewed RCTs published in English between 2000 and 2020 examining the effect of PA on motor skill development in healthy children. Results. A total of 25 RCTs were included, 20 (80%) of which reported significant improvements in children's motor skill performance. Specifically, 18 studies examined traditional PA interventions and 7 studies examined exergaming-based PA interventions, 83% and 71% of which observed statistically significant improvements in children's motor skill development, respectively. Conclusions. Findings support the causal evidence regarding the effects of PA on motor skill development in children. Notable limitations of this review included heterogeneity of measurement protocols and assessment tools used to test children's motor skills across studies, a wide range of PA intervention dose across studies, and the lack of power analyses and long-term follow-up assessments in individual studies to discern appropriate sample sizes and long-term effectiveness, respectively. To further strengthen the evidence in this emerging field, we advocate for future RCTs to employ a priori power analyses, long-term follow-up measurements, and more exergaming-based interventions to allow for comparisons with traditional PA interventions, to explore the dose response and moderating relationships between PA and motor skill development in childhood, and to utilize homogenous assessment instruments to allow for more rigorous, quantitative syntheses. [ABSTRACT FROM AUTHOR]

Existing video salient object detection (VSOD) methods focus on exploring either short-term or long-term temporal information. However, temporal information is exploited in a global frame-level or regular grid structure, neglecting inter-frame structural dependencies. In this article, we propose to learn long-term structural dependencies with a structure-evolving graph convolutional network (GCN). Particularly, we construct a graph for the entire video using a fast supervoxel segmentation method, in which each node is connected according to spatio-temporal structural similarity. We infer the inter-frame structural dependencies of salient object using convolutional operations on the graph. To prune redundant connections in the graph and better adapt to the moving salient object, we present an adaptive graph pooling to evolve the structure of the graph by dynamically merging similar nodes, learning better hierarchical representations of the graph. Experiments on six public datasets show that our method outperforms all other state-of-the-art methods. Furthermore, We also demonstrate that our proposed adaptive graph pooling can effectively improve the supervoxel algorithm in the term of segmentation accuracy. [ABSTRACT FROM AUTHOR]

Supervoxel segmentation algorithm has been applied as a preprocessing step for many vision tasks. However, existing supervoxel segmentation algorithms cannot generate hierarchical supervoxel segmentation well preserving the spatiotemporal boundaries in real time, which prevents the downstream applications from accurate and efficient processing. In this paper, we propose a real-time hierarchical supervoxel segmentation algorithm based on the minimum spanning tree (MST), which achieves state-of-the-art accuracy meanwhile at least $11\times $ faster than existing methods. In particular, we present a dynamic graph updating operation into the iterative construction process of the MST, which can geometrically decrease the numbers of vertices and edges. In this way, the proposed method is able to generate arbitrary scales of supervoxels on the fly. We prove the efficiency of our algorithm that can produce hierarchical supervoxels in the time complexity of $O(n)$ , where $n$ denotes the number of voxels in the input video. Quantitative and qualitative evaluations on public benchmarks demonstrate that our proposed algorithm significantly outperforms the state-of-the-art algorithms in terms of supervoxel segmentation accuracy and computational efficiency. Furthermore, we demonstrate the effectiveness of the proposed method on a downstream application of video object segmentation. [ABSTRACT FROM AUTHOR]

Mobile devices usually mount a depth sensor to resolve ill-posed problems, like salient object detection on cluttered background. The main barrier of exploring RGBD data is to handle the information from two different modalities. To cope with this problem, in this paper, we propose a boundary-aware cross-modal fusion network for RGBD salient object detection. In particular, to enhance the fusion of color and depth features, we present a cross-modal feature sampling module to balance the contribution of the RGB and depth features based on the statistics of their channel values. In addition, in our multi-scale dense fusion network architecture, we not only incorporate edge-sensitive losses to preserve the boundary of the detected salient region, but also refine its structure by merging the estimated saliency maps of different scales. We accomplish the multi-scale saliency map merging using two alternative methods which produce refined saliency maps via per-pixel weighted combination and an encoder-decoder network. Extensive experimental evaluations demonstrate that our proposed framework can achieve the state-of-the-art performance on several public RGBD-based datasets. [ABSTRACT FROM AUTHOR]

Crowd counting is challenging due to unconstrained imaging factors, e.g., background clutters, non-uniform distribution of people, large scale and perspective variations. Dealing with these problems using deep neural networks requires rich prior knowledge and multi-scale contextual representations. In this paper, we propose a Cross-stage Refinement Network (CRNet) that can refine predicted density maps progressively based on hierarchical multi-level density priors. In particular, CRNet is composed of several fully convolutional networks. They are stacked together recursively with the previous output as the next input, and each of them serves to utilize previous density output to gradually correct prediction errors of crowd areas and refine the predicted density maps at different stages. Cross-stage multi-level density priors are further exploited in our recurrent framework by the cross-stage skip layers based on ConvLSTM. To cope with different challenges of unconstrained crowd scenes, we explore different crowd-specific data augmentation methods to mimic real-world scenarios and enrich crowd feature representations from different aspects. Extensive experiments show the proposed method achieves superior performances against state-of-the-art methods on four widely-used challenging benchmarks in terms of counting accuracy and density map quality. Code and models are available at this https://github.com/lytgftyf/Crowd-Counting-via-Cross-stage-Refinement-Networks. [ABSTRACT FROM AUTHOR]

In this paper, we propose a novel method to generate stereoscopic images from light-field images with the intended depth range and simultaneously perform image super-resolution. Subject to the small baseline of neighboring subaperture views and low spatial resolution of light-field images captured using compact commercial light-field cameras, the disparity range of any two subaperture views is usually very small. We propose a method to control the disparity range of the target stereoscopic images with linear or nonlinear disparity scaling and properly resolve the disocclusion problem with the aid of a smooth energy term previously used for texture synthesis. The left and right views of the target stereoscopic image are simultaneously generated by a unified optimization framework, which preserves content coherence between the left and right views by a coherence energy term. The disparity range of the target stereoscopic image can be larger than that of the input light field image. This benefits many light field image-based applications, e.g., displaying light field images on various stereo display devices and generating stereoscopic panoramic images from a light field image montage. An extensive experimental evaluation demonstrates the effectiveness of our method. [ABSTRACT FROM AUTHOR]

Developing a safe and efficient collision-avoidance policy for multiple robots is challenging in the decentralized scenarios where each robot generates its paths with limited observation of other robots' states and intentions. Prior distributed multi-robot collision-avoidance systems often require frequent inter-robot communication or agent-level features to plan a local collision-free action, which is not robust and computationally prohibitive. In addition, the performance of these methods is not comparable with their centralized counterparts in practice. In this article, we present a decentralized sensor-level collision-avoidance policy for multi-robot systems, which shows promising results in practical applications. In particular, our policy directly maps raw sensor measurements to an agent's steering commands in terms of the movement velocity. As a first step toward reducing the performance gap between decentralized and centralized methods, we present a multi-scenario multi-stage training framework to learn an optimal policy. The policy is trained over a large number of robots in rich, complex environments simultaneously using a policy-gradient-based reinforcement-learning algorithm. The learning algorithm is also integrated into a hybrid control framework to further improve the policy's robustness and effectiveness. We validate the learned sensor-level collision-3avoidance policy in a variety of simulated and real-world scenarios with thorough performance evaluations for large-scale multi-robot systems. The generalization of the learned policy is verified in a set of unseen scenarios including the navigation of a group of heterogeneous robots and a large-scale scenario with 100 robots. Although the policy is trained using simulation data only, we have successfully deployed it on physical robots with shapes and dynamics characteristics that are different from the simulated agents, in order to demonstrate the controller's robustness against the simulation-to-real modeling error. Finally, we show that the collision-avoidance policy learned from multi-robot navigation tasks provides an excellent solution for safe and effective autonomous navigation for a single robot working in a dense real human crowd. Our learned policy enables a robot to make effective progress in a crowd without getting stuck. More importantly, the policy has been successfully deployed on different types of physical robot platforms without tedious parameter tuning. Videos are available at https://sites.google.com/view/hybridmrca. [ABSTRACT FROM AUTHOR]

With the advances in commercial light field cameras, light field image processing has attracted considerable attention from researchers. In this paper, we propose a novel method for generating stereoscopic images from a light field image pair with flexible control over the convergence of the virtual stereo cameras. We have developed a light field image-capturing prototype that consists of two horizontally arranged light field cameras (i.e., with their optical axes being parallel to each other). When using our proposed device for image/video capture, stereo photographers can concentrate on how to capture the desired visual experience without being frequently disturbed by having to manipulate the stereo camera parameters, i.e., the convergence angle of a stereo camera. During postprocessing, our method estimates accurate disparity maps for the light field image pair and then generates the target stereoscopic images that satisfy the desired stereo camera convergence requirements by adopting a novel view synthesis method for light field images. We have conducted extensive experiments to demonstrate the effectiveness of our proposed method. [ABSTRACT FROM AUTHOR]