Dataset Information

Robust action recognition using multi-scale spatial-temporal concatenations of local features as natural action structures.

ABSTRACT: Human and many other animals can detect, recognize, and classify natural actions in a very short time. How this is achieved by the visual system and how to make machines understand natural actions have been the focus of neurobiological studies and computational modeling in the last several decades. A key issue is what spatial-temporal features should be encoded and what the characteristics of their occurrences are in natural actions. Current global encoding schemes depend heavily on segmenting while local encoding schemes lack descriptive power. Here, we propose natural action structures, i.e., multi-size, multi-scale, spatial-temporal concatenations of local features, as the basic features for representing natural actions. In this concept, any action is a spatial-temporal concatenation of a set of natural action structures, which convey a full range of information about natural actions. We took several steps to extract these structures. First, we sampled a large number of sequences of patches at multiple spatial-temporal scales. Second, we performed independent component analysis on the patch sequences and classified the independent components into clusters. Finally, we compiled a large set of natural action structures, with each corresponding to a unique combination of the clusters at the selected spatial-temporal scales. To classify human actions, we used a set of informative natural action structures as inputs to two widely used models. We found that the natural action structures obtained here achieved a significantly better recognition performance than low-level features and that the performance was better than or comparable to the best current models. We also found that the classification performance with natural action structures as features was slightly affected by changes of scale and artificially added noise. We concluded that the natural action structures proposed here can be used as the basic encoding units of actions and may hold the key to natural action understanding.

SUBMITTER: Zhu X

PROVIDER: S-EPMC3464264 | biostudies-literature | 2012

REPOSITORIES: biostudies-literature

ACCESS DATA

Publications

Robust action recognition using multi-scale spatial-temporal concatenations of local features as natural action structures.

Zhu Xiaoyuan X Li Meng M Li Xiaojian X Yang Zhiyong Z Tsien Joe Z JZ

PloS one 20121004 10

Human and many other animals can detect, recognize, and classify natural actions in a very short time. How this is achieved by the visual system and how to make machines understand natural actions have been the focus of neurobiological studies and computational modeling in the last several decades. A key issue is what spatial-temporal features should be encoded and what the characteristics of their occurrences are in natural actions. Current global encoding schemes depend heavily on segmenting w ...[more]

PMID: 23056403

Similar Datasets

Project description:BackgroundDengue is a vector-borne disease caused by the dengue virus (DENV). Despite the crucial role of Aedes mosquitoes in DENV transmission, pure vector indices poorly correlate with human infections. Therefore there is great need for a better understanding of the spatial and temporal scales of DENV transmission between mosquitoes and humans. Here, we have systematically monitored the circulation of DENV in individual Aedes spp. mosquitoes and human patients from Caratinga, a dengue endemic city in the state of Minas Gerais, in Southeast Brazil. From these data, we have developed a novel stochastic point process pattern algorithm to identify the spatial and temporal association between DENV infected mosquitoes and human patients.MethodsThe algorithm comprises of: (i) parameterization of the variogram for the incidence of each DENV serotype in mosquitoes; (ii) identification of the spatial and temporal ranges and variances of DENV incidence in mosquitoes in the proximity of humans infected with dengue; and (iii) analysis of the association between a set of environmental variables and DENV incidence in mosquitoes in the proximity of humans infected with dengue using a spatio-temporal additive, geostatistical linear model.ResultsDENV serotypes 1 and 3 were the most common virus serotypes detected in both mosquitoes and humans. Using the data on each virus serotype separately, our spatio-temporal analyses indicated that infected humans were located in areas with the highest DENV incidence in mosquitoes, when incidence is calculated within 2.5-3 km and 50 days (credible interval 30-70 days) before onset of symptoms in humans. These measurements are in agreement with expected distances covered by mosquitoes and humans and the time for virus incubation. Finally, DENV incidence in mosquitoes found in the vicinity of infected humans correlated well with the low wind speed, higher air temperature and northerly winds that were more likely to favor vector survival and dispersal in Caratinga.ConclusionsWe have proposed a new way of modeling bivariate point pattern on the transmission of arthropod-borne pathogens between vector and host when the location of infection in the latter is known. This strategy avoids some of the strong and unrealistic assumptions made by other point-process models. Regarding virus transmission in Caratinga, our model showed a strong and significant association between high DENV incidence in mosquitoes and the onset of symptoms in humans at specific spatial and temporal windows. Together, our results indicate that vector surveillance must be a priority for dengue control. Nevertheless, localized vector control at distances lower than 2.5 km around premises with infected vectors in densely populated areas are not likely to be effective.

Project description:BackgroundDiseases transmitted by invasive Aedes aegypti and Aedes albopictus mosquitoes are public health issues in the tropics and subtropics. Understanding the ecology of mosquito vectors is essential for the development of effective disease mitigation programs and will allow for accurate predictions of vector occurrence and abundance. Studies that examine mosquito population dynamics are typically focused on female presence or total adult captures without discriminating the temporal and spatial distribution of both sexes.MethodsWe collected immature and adult mosquitoes bimonthly for 2 years (2018-2019) in the Medellín Botanical Garden. Collection sites differed in proximity to buildings and nearby vegetation, and were classified by their overhead vegetation cover. We used linear mixed models (LMMs) and Spatial Analysis by Distance Indices (SADIE) to assess the spatial distribution of Ae. aegypti and Ae. albopictus. Using our Ae. albopictus captures exclusively, we assessed (1) the spatial and temporal distribution of males and females using SADIE and a generalized linear mixed model (GLMM), (2) the relationship between climatic variables/vegetation coverage and adult captures using GLMMs and LMMs, and (3) the correlation of male and female size in relation to climatic variables and vegetation coverage using LMMs.ResultsSpatial analysis showed that Ae. aegypti and Ae. albopictus were distributed at different locations within the surveilled area. However, Ae. albopictus was the predominant species in the park during the study period. Adult Ae. albopictus captures were positively correlated with precipitation and relative humidity, and inversely correlated with temperature and wind speed. Moreover, we observed a spatial misalignment of Ae. albopictus males and females-the majority of males were located in the high vegetation coverage sites, while females were more evenly distributed. We observed significant associations of the size of our adult Ae. albopictus captures with precipitation, temperature, and wind speed for both sexes and found that overhead vegetation cover influenced male size, but observed no effect on female size.ConclusionsOur work elucidates the differential dynamics of Ae. albopictus males and females, which is pivotal to develop accurate surveillance and the successful establishment of vector control programs based on the disruption of insect reproduction.

Dataset Information

Robust action recognition using multi-scale spatial-temporal concatenations of local features as natural action structures.

Publications

Robust action recognition using multi-scale spatial-temporal concatenations of local features as natural action structures.

Similar Datasets

OmicsDI is part of the ELIXIR infrastructure

Tweets