Recent research on super-resolution has progressed with the development of deep convolutional neural networks (DCNN). In particular, residual learning techniques exhibit improved performance. In this paper, we develop an enhanced deep super-resolution network (EDSR) with performance exceeding those of current state-of-the-art SR methods. The significant performance improvement of our model is due to optimization by removing unnecessary modules in conventional residual networks. The performance is further improved by expanding the model size while we stabilize the training procedure. We also propose a new multi-scale deep super-resolution system (MDSR) and training method, which can reconstruct high-resolution images of different upscaling factors in a single model. The proposed methods show superior performance over the state-of-the-art methods on benchmark datasets and prove its excellence by winning the NTIRE2017 Super-Resolution Challenge [26].

No abstract

Prevailing video frame interpolation techniques rely heavily on optical flow estimation and require additional model complexity and computational cost; it is also susceptible to error propagation in challenging scenarios with large motion and heavy occlusion. To alleviate the limitation, we propose a simple but effective deep neural network for video frame interpolation, which is end-to-end trainable and is free from a motion estimation network component. Our algorithm employs a special feature reshaping operation, referred to as PixelShuffle, with a channel attention, which replaces the optical flow computation module. The main idea behind the design is to distribute the information in a feature map into multiple channels and extract motion information by attending the channels for pixel-level frame synthesis. The model given by this principle turns out to be effective in the presence of challenging motion and occlusion. We construct a comprehensive evaluation benchmark and demonstrate that the proposed approach achieves outstanding performance compared to the existing models with a component for optical flow computation.

No abstract

This study investigates the effects of mHealth interventions on sustainable behavior change and weight loss, drawing on in-app user activity data and online survey data. Specifically, we focus on the interactions within mobile support groups in Noom, an mHealth application for obesity intervention, to delve into how social support from facilitators and peers may play differential roles in promoting health outcomes. The results of structural equation modeling (N = 301) demonstrated that (a) perceived facilitator support was positively associated with group members' health information acquisition such as fitness-themed article reading whereas perceived peer support was positively linked to group participation such as posting and responding; (b) perceived peer support was positively related to normative influence among group members, which subsequently increased group members' responses to others' posts; and (c) health information reading and in-group posting promoted weight loss; however, merely responding to others' posts did not lead to weight-loss success. The findings suggest that the complementary influences of facilitators and peers must be considered to enhance the efficacy of support group interventions.

Despite the growing popularity of mHealth applications, their usage outcomes have received limited empirical attention. Drawing on server-level user activity data and an online survey (N = 384), this study examines the use of an mHealth application for weight loss to elucidate the ways in which it can help individuals harness the power of self-efficacy and group support to enact behavior change and accomplish their health goals. The results of structural equation modeling based on 6-month user activity data demonstrated that (a) self-efficacy had a positive impact on persistent food logging in an mHealth application; (b) social support received from a mobile group was positively associated with food logging and group participation; and (c) both food logging and group participation predicted weight loss success. Extending these findings, this study suggests theoretical and practical implications for designing individually tailored and evidence-based health intervention strategies using advanced mHealth technologies.

PurposeThis study was performed to investigate the effectiveness of gastric cancer (GC) screening methods in a community-based prospective cohort of the Korean Multi-center Cancer Cohort (KMCC) with over a 10-year follow-up.Materials and MethodsA total 10,909 and 4,773 subjects from the KMCC with information on gastroendoscopy (GE) and upper gastrointestinal series (UGIS) were included in this study. Cox proportional hazard model adjusted for age, sex, Helicobacter pylori infection, cigarette smoking, and alcohol drinking was used to estimate the hazard ratios (HRs) and 95% confidence interval (CI).ResultsThe GE screened subjects had almost half the risk of GC-specific death than that of unscreened subjects (HR, 0.58; 95% CI, 0.36 to 0.94). Among the GC patients, GE screenees had a 2.24-fold higher survival rate than that of the non-screenees (95% CI, 1.61 to 3.11). In particular, GE screenees who underwent two or more screening episodes had a higher survival rate than that of the non-screenees (HR, 13.11; 95% CI, 7.38 to 23.30). The effectiveness of GE screening on reduced GC mortality and increased survival rate of GC patients was better in elderly subjects (≥ 65 years old) (HR, 0.47; 95% CI, 0.24 to 0.95 and HR, 8.84; 95% CI, 3.63 to 21.57, respectively) than that in younger subjects (< 65 years old) (HR, 0.66; 95% CI, 0.34 to 1.29 and HR, 1.83; 95% CI, 1.24 to 2.68, respectively). In contrast, UGIS screening had no significant relation to GC mortality and survival.ConclusionThe findings of this study suggest that a decreased GC-specific mortality and improved survival rate in GC patients can be achieved through GE screening.

Research on online communities has emphasized the individual benefits of social support for members, but less is known about how such communities are regulated through organizing processes of support and control. Drawing on a survey of 214 members of a particular online message board community, we develop and test a model of social support, strength of ties, normative influence, and concertive control and their influence on members’ sense of virtual community (SOVC). We find that all four factors predict SOVC, but that normative influence and concertive control have the strongest effects. Furthermore, social support and concertive control mediate the effects of number of strong ties and normative influence (respectively) on SOVC. Finally, we find no association between SOVC and time-lagged posting frequency. Our findings have important implications for understanding the factors that lead to attachment in online communities, and they suggest that sense of belonging works through tandem communicative processes of support and control.