A Channel-Wise Spatial-Temporal Aggregation Network for Action Recognition

A very challenging task for action recognition concerns how to effectively extract and utilize the temporal and spatial information of video (especially temporal information). To date, many researchers have proposed various spatial-temporal convolution structures. Despite their success, most models are limited in further performance especially on those datasets that are highly time-dependent due to their failure to identify the fusion relationship between the spatial and temporal features inside the convolution channel. In this paper, we proposed a lightweight and efficient spatial-temporal extractor, denoted as Channel-Wise Spatial-Temporal Aggregation block (CSTA block), which could be flexibly plugged in existing 2D CNNs (denoted by CSTANet). The CSTA Block utilizes two branches to model spatial-temporal information separately. In temporal branch, It is equipped with a Motion Attention Module (MA), which is used to enhance the motion regions in a given video. Then, we introduced a Spatial-Temporal Channel Attention (STCA) module, which could aggregate spatial-temporal features of each block channel-wisely in a self-adaptive and trainable way. The final experimental results demonstrate that the proposed CSTANet achieved the state-of-the-art results on EGTEA Gaze++ and Diving48 datasets, and obtained competitive results on Something-Something V1&V2 at the less computational cost.

Epidemiological features and spatial–temporal distribution of visceral leishmaniasis in mainland China: a population-based surveillance study from 2004 to 2019

Background Although visceral leishmaniasis (VL) was largely brought under control in most regions of China during the previous century, VL cases have rebounded in western and central China in recent decades. The aim of this study was to investigate the epidemiological features and spatial–temporal distribution of VL in mainland China from 2004 to 2019. Methods Incidence and mortality data for VL during the period 2004–2019 were collected from the Public Health Sciences Data Center of China and annual national epidemic reports of VL, whose data source was the National Diseases Reporting Information System. Joinpoint regression analysis was performed to explore the trends of VL. Spatial autocorrelation and spatial–temporal clustering analysis were conducted to identify the distribution and risk areas of VL transmission. Results A total of 4877 VL cases were reported in mainland China during 2004–2019, with mean annual incidence of 0.0228/100,000. VL incidence showed a decreasing trend in general during our study period (annual percentage change [APC] = −4.2564, 95% confidence interval [CI]: −8.0856 to −0.2677). Among mainly endemic provinces, VL was initially heavily epidemic in Gansu, Sichuan, and especially Xinjiang, but subsequently decreased considerably. In contrast, Shaanxi and Shanxi witnessed significantly increasing trends, especially in 2017–2019. The first-level spatial–temporal aggregation area covered two endemic provinces in northwestern China, including Gansu and Xinjiang, with the gathering time from 2004 to 2011 (relative risk [RR] = 13.91, log-likelihood ratio [LLR] = 3308.87, P < 0.001). The secondary aggregation area was detected in Shanxi province of central China, with the gathering time of 2019 (RR = 1.61, LLR = 4.88, P = 0.041). The epidemic peak of October to November disappeared in 2018–2019, leaving only one peak in March to May. Conclusions Our findings suggest that VL is still an important endemic infectious disease in China. Epidemic trends in different provinces changed significantly and spatial–temporal aggregation areas shifted from northwestern to central China during our study period. Mitigation strategies, including large-scale screening, insecticide spraying, and health education encouraging behavioral change, in combination with other integrated approaches, are needed to decrease transmission risk in areas at risk, especially in Shanxi, Shaanxi, and Gansu provinces. Graphical abstract

Inter-Temporal Aggregation for Spatially Explicit Optimal Harvest Scheduling under Area Restrictions

We propose a new approach to solve inter-temporal unit aggregation issues under maximum opening size requirements using two models. The first model is based on Model I formulation with static harvest treatments for harvest activities. This model identifies periodic harvest activities using a set of constraints for inter-temporal aggregation. The second model is based on Model II formulation, which uses dynamic harvest treatments and incorporates periodic harvest activities directly into the model formulation. The proposed approach contributes to the literature on spatially constrained harvest scheduling problems as it allows a pattern of unit aggregation to change across multiple harvests over time, as inter-temporal aggregation under a maximum opening size requirement over period-specific duration. The main idea of the proposed approach for inter-temporal aggregation is to use a multiple layer scheme for a set of spatial constraints, which is adapted from a maximum flow specification in a spatial forest unit network and a sequential triangle connection to create fully connected feasible clusters. By dividing the planning horizon into period-specific durations for different spatial aggregation patterns, the models can complete inter-temporal spatial aggregation over the planning horizon under a maximum opening size requirement per duration. Study Implications Inter-temporal unit aggregation is important because it provides flexible aggregation patterns for maximum opening size problems with multiple harvests over time. We have proposed a new modeling approach capable of solving spatially constrained harvest scheduling problems by allowing a pattern of unit aggregation to change across multiple harvest periods over time, as inter-temporal aggregation under flexible maximum opening size requirements. Forest managers can benefit from this approach for their future requirements based on the public interests as well as their own.

Analyzing Travel Time Reliability from Sparse Probe Vehicle Data: A Case Study on the Effects of Spatial and Temporal Aggregation

Travel time reliability (TTR) indices have gained considerable attention for evaluating the quality of traffic infrastructure. Whereas TTR measures have been widely explored using data from stationary sensors with high penetration rates, there is a lack of research on calculating TTR from mobile sensors such as probe vehicle data (PVD) which is characterized by low penetration rates. PVD is a relevant data source for analyzing non-highway routes, as they are often not sufficiently covered by stationary sensors. The paper presents a methodology for analyzing TTR on (sub-)urban and rural routes with sparse PVD as the only data source that could be used by road authorities or traffic planners. Especially in the case of sparse data, spatial and temporal aggregations could have great impact, which are investigated on two levels: first, the width of time of day (TOD) intervals and second, the length of road segments. The spatial and temporal aggregation effects on travel time index (TTI) as prominent TTR measure are analyzed within an exemplary case study including three different routes. TTI patterns are calculated from data of one year grouped by different days-of-week (DOW) groups and the TOD. The case study shows that using well-chosen temporal and spatial aggregations, even with sparse PVD, an in-depth analysis of traffic patterns is possible.

Epidemiological Trends and Hotspots of Other Infectious Diarrhea (OID) in Mainland China: A Population-Based Surveillance Study From 2004 to 2017

Background: The incidence of other infectious diarrhea (OID) ranked second in class C notifiable disease in China. It has posed a great threat to public health of all age groups. The aim of this study was to investigate the epidemiological trends and hotspots of OID in mainland China.Materials and Methods: Incidence and mortality data for OID stratified by date, age and region from 2004 to 2017 was extracted from the data-center of China public health science. Joinpoint regression and space-time analyses were performed to explore the epidemiological trends and hotspots of OID.Results: The average annual incidence of OID was 60.64/100,000 and it showed an increased trend in the mainland China especially after 2006 (APC = 4.12, 95 CI%: 2.06–6.21). Children of 0–4 year age group accounts for 60.00% (5,820,897/11,414,247) of all cases and its incidence continuously increased though 2004–2017 (APC = 6.65, 95 CI%: 4.39–8.96). The first-level spatial and temporal aggregation areas were located in Beijing and Tianjin, with the gathering time from 2005/1/1 to 2011/12/31 (RR = 5.52, LLR = 572893.59, P &lt; 0.001). The secondary spatial and temporal aggregation areas covered Guangdong, Guangxi, Hainan and Guizhou from 2011/1/1 to 2017/12/31 (RR = 1.98, LLR = 242292.72, P &lt; 0.001). OID of Tianjin and Beijing presented a decreased trend since 2006. However, the incidence of OID in Guangdong, Guangxi, Hainan and Guizhou showed increased trends through 2004–2017.Conclusion: Our study showed that OID showed a constantly increasing trend and brought considerable burden in China especially in the 0–4 age group. The high-risk periods and clusters of regions for OID were identified, which will help government develop disease-specific and location-specific interventive measures.