The Experts below are selected from a list of 167265 Experts worldwide ranked by ideXlab platform
Yuan Yuan - One of the best experts on this subject based on the ideXlab platform.
-
AAAI - Memory-Augmented Temporal Dynamic Learning for Action Recognition
Proceedings of the AAAI Conference on Artificial Intelligence, 2019Co-Authors: Yuan Yuan, Dong Wang, Qi WangAbstract:Human actions captured in video sequences contain two crucial factors for action recognition, i.e., visual appearance and motion Dynamics. To model these two aspects, Convolutional and Recurrent Neural Networks (CNNs and RNNs) are adopted in most existing successful methods for recognizing actions. However, CNN based methods are limited in modeling long-term motion Dynamics. RNNs are able to learn Temporal motion Dynamics but lack effective ways to tackle unsteady Dynamics in long-duration motion. In this work, we propose a memory-augmented Temporal Dynamic learning network, which learns to write the most evident information into an external memory module and ignore irrelevant ones. In particular, we present a differential memory controller to make a discrete decision on whether the external memory module should be updated with current feature. The discrete memory controller takes in the memory history, context embedding and current feature as inputs and controls information flow into the external memory module. Additionally, we train this discrete memory controller using straight-through estimator. We evaluate this end-to-end system on benchmark datasets (UCF101 and HMDB51) of human action recognition. The experimental results show consistent improvements on both datasets over prior works and our baselines.
-
Memory-Augmented Temporal Dynamic Learning for Action Recognition
arXiv: Computer Vision and Pattern Recognition, 2019Co-Authors: Yuan Yuan, Dong Wang, Qi WangAbstract:Human actions captured in video sequences contain two crucial factors for action recognition, i.e., visual appearance and motion Dynamics. To model these two aspects, Convolutional and Recurrent Neural Networks (CNNs and RNNs) are adopted in most existing successful methods for recognizing actions. However, CNN based methods are limited in modeling long-term motion Dynamics. RNNs are able to learn Temporal motion Dynamics but lack effective ways to tackle unsteady Dynamics in long-duration motion. In this work, we propose a memory-augmented Temporal Dynamic learning network, which learns to write the most evident information into an external memory module and ignore irrelevant ones. In particular, we present a differential memory controller to make a discrete decision on whether the external memory module should be updated with current feature. The discrete memory controller takes in the memory history, context embedding and current feature as inputs and controls information flow into the external memory module. Additionally, we train this discrete memory controller using straight-through estimator. We evaluate this end-to-end system on benchmark datasets (UCF101 and HMDB51) of human action recognition. The experimental results show consistent improvements on both datasets over prior works and our baselines.
-
Temporal Dynamic Graph LSTM for Action-driven Video Object Detection
arXiv: Computer Vision and Pattern Recognition, 2017Co-Authors: Yuan Yuan, Xiaodan Liang, Xiaolong Wang, Dit-yan Yeung, Abhinav GuptaAbstract:In this paper, we investigate a weakly-supervised object detection framework. Most existing frameworks focus on using static images to learn object detectors. However, these detectors often fail to generalize to videos because of the existing domain shift. Therefore, we investigate learning these detectors directly from boring videos of daily activities. Instead of using bounding boxes, we explore the use of action descriptions as supervision since they are relatively easy to gather. A common issue, however, is that objects of interest that are not involved in human actions are often absent in global action descriptions known as "missing label". To tackle this problem, we propose a novel Temporal Dynamic graph Long Short-Term Memory network (TD-Graph LSTM). TD-Graph LSTM enables global Temporal reasoning by constructing a Dynamic graph that is based on Temporal correlations of object proposals and spans the entire video. The missing label issue for each individual frame can thus be significantly alleviated by transferring knowledge across correlated objects proposals in the whole video. Extensive evaluations on a large-scale daily-life action dataset (i.e., Charades) demonstrates the superiority of our proposed method. We also release object bounding-box annotations for more than 5,000 frames in Charades. We believe this annotated data can also benefit other research on video-based object recognition in the future.
-
ICCV - Temporal Dynamic Graph LSTM for Action-Driven Video Object Detection
2017 IEEE International Conference on Computer Vision (ICCV), 2017Co-Authors: Yuan Yuan, Xiaodan Liang, Xiaolong Wang, Dit-yan Yeung, Abhinav GuptaAbstract:In this paper, we investigate a weakly-supervised object detection framework. Most existing frameworks focus on using static images to learn object detectors. However, these detectors often fail to generalize to videos because of the existing domain shift. Therefore, we investigate learning these detectors directly from boring videos of daily activities. Instead of using bounding boxes, we explore the use of action descriptions as supervision since they are relatively easy to gather. A common issue, however, is that objects of interest that are not involved in human actions are often absent in global action descriptions known as “missing label”. To tackle this problem, we propose a novel Temporal Dynamic graph Long Short-Term Memory network (TDGraph LSTM). TD-Graph LSTM enables global Temporal reasoning by constructing a Dynamic graph that is based on Temporal correlations of object proposals and spans the entire video. The missing label issue for each individual frame can thus be significantly alleviated by transferring knowledge across correlated objects proposals in the whole video. Extensive evaluations on a large-scale daily-life action dataset (i.e., Charades) demonstrates the superiority of our proposed method. We also release object bounding-box annotations for more than 5,000 frames in Charades. We believe this annotated data can also benefit other research on video-based object recognition in the future.
Guangjin Tian - One of the best experts on this subject based on the ideXlab platform.
-
The spatio-Temporal Dynamic pattern of rural domestic solid waste discharge of China and its challenges.
Environmental science and pollution research international, 2018Co-Authors: Guangjin Tian, Lingqiang Kong, Xiaojuan Liu, Wenping YuanAbstract:At present, construction of rural domestic waste treatment facilities is seriously lagging, and in many cases, treatment facilities do not yet exist in some villages of China. Serious rural waste pollution has not only impacted the quality of surface water and groundwater but also the atmosphere and the living environment of farmers of China. There are relatively few studies of rural domestic waste pollution, especially with respect to the spatio-Temporal Dynamic pattern of rural domestic waste discharge. Using survey data and income per capita, we calculated rural domestic waste discharge per capita per day. From this, we calculated provincial rural domestic waste discharge. According to our study, rural domestic waste discharge was 1.42 × 108 t/year in 2000. This number increased to 2.3 × 108 t/year in 2006 and to 2.47 × 108 t/year in 2010. Rural domestic waste increased dramatically while the actual rural population and the proportion of the rural population declined. When examining the eight regions, the rural domestic waste discharge of northeastern China, Qinghai-Tibet, middle China, and southwestern China had increased dramatically, while that of northern China, southern China, and eastern China increased relatively slowly. The economies of northern China, southern China, and eastern China are more developed; their rural domestic waste discharge has been high since 2000 and has continued to increase slowly. In northeastern China, Qinghai-Tibet, middle China, and southwestern China, rural domestic waste discharge was low in 2000; however, in the ten-year period from 2000 to 2010, their rural domestic waste discharge increased dramatically.
-
the urban growth size distribution and spatio Temporal Dynamic pattern of the yangtze river delta megalopolitan region china
Ecological Modelling, 2011Co-Authors: Guangjin Tian, Jing Jiang, Zhifeng Yang, Yaoqi ZhangAbstract:As one of the six megalopolitan regions in the world, the Yangtze River Delta is one of the most populated and developed regions of China. The spatial and Temporal Dynamic pattern of the urbanization process of the megalopolitan region is investigated. This work compared the spatial and Temporal Dynamic pattern of the urban growth for the five urban areas (Shanghai, Nanjing, Suzhou, Wuxi and Changzhou) in this region. During the 15 years, urban growth patterns were dramatically uneven over three 5-year periods. The size distribution of the five urban areas became more even with the rapid urbanization process. The patterns of urban expansion reflected policy adjustment and economic development throughout the time. Landscape metric analysis across concentric buffer zones was conducted to elucidate the area, shape, size, complexity and configuration of urban expansion. The study indicates the coalescence process occurred during the rapid urban growth from 1990 to 1995 and the moderate growth period from 2000 to 2005, but different urban growth period between 1995 and 2000. The urban growth pattern was coalesced for the Nanjing and Wuxi metropolitan areas and diffused for Shanghai, Suzhou and Changzhou. This approach indicates that the coalescence process was the major growth model for this region in the recent 15 years despite their different size, economic growth and population growth. The diffusion-coalesce dichotomy represent endpoints rather than alternate states of urban growth. This work will be beneficial in understanding the size distribution and urbanization process of the megalopolitan region in China.
-
The spatio-Temporal Dynamic pattern of rural residential land in China in the 1990s using Landsat TM images and GIS.
Environmental management, 2007Co-Authors: Guangjin Tian, Zhifeng Yang, Yaoqi ZhangAbstract:Through interpreting Landsat TM images, this study analyzes the spatial distribution of rural settlements in China in 2000. It calculates rural residential land percentage for every 1-km(2) cell. The entire country is divided into 33 regions to investigate the spatio-Temporal Dynamic patterns of rural residential land during the 1990s. According to the remote sensing survey, the rural residential land increased by 7.88 x 10(5) ha in the 1990s. The increment of rural residential land was 0.55 million ha in 1990-1995 and 0.23 million ha in 1995-2000. In 1990-1995, rural residential land increased dramatically in the eastern regions such as the Yangtze River Delta, Pearl River Delta, and North China Plain, accounting for 80.80% of the national growth; the expansion in the western regions was much more moderate. In 1995-2000, the expansion of rural residential land in eastern regions slowed, accounting for only 58.54% of the increase at the national level, whereas the expansion in the western regions accelerated. Rapid rural residential development resulted from increasing home construction and the limited control on rural land. The great regional disparity reflected the regional economic development and land-use policy change. Our finding shows that nearly 60% of the rural residential area came from cropland.
Yaoqi Zhang - One of the best experts on this subject based on the ideXlab platform.
-
the urban growth size distribution and spatio Temporal Dynamic pattern of the yangtze river delta megalopolitan region china
Ecological Modelling, 2011Co-Authors: Guangjin Tian, Jing Jiang, Zhifeng Yang, Yaoqi ZhangAbstract:As one of the six megalopolitan regions in the world, the Yangtze River Delta is one of the most populated and developed regions of China. The spatial and Temporal Dynamic pattern of the urbanization process of the megalopolitan region is investigated. This work compared the spatial and Temporal Dynamic pattern of the urban growth for the five urban areas (Shanghai, Nanjing, Suzhou, Wuxi and Changzhou) in this region. During the 15 years, urban growth patterns were dramatically uneven over three 5-year periods. The size distribution of the five urban areas became more even with the rapid urbanization process. The patterns of urban expansion reflected policy adjustment and economic development throughout the time. Landscape metric analysis across concentric buffer zones was conducted to elucidate the area, shape, size, complexity and configuration of urban expansion. The study indicates the coalescence process occurred during the rapid urban growth from 1990 to 1995 and the moderate growth period from 2000 to 2005, but different urban growth period between 1995 and 2000. The urban growth pattern was coalesced for the Nanjing and Wuxi metropolitan areas and diffused for Shanghai, Suzhou and Changzhou. This approach indicates that the coalescence process was the major growth model for this region in the recent 15 years despite their different size, economic growth and population growth. The diffusion-coalesce dichotomy represent endpoints rather than alternate states of urban growth. This work will be beneficial in understanding the size distribution and urbanization process of the megalopolitan region in China.
-
The spatio-Temporal Dynamic pattern of rural residential land in China in the 1990s using Landsat TM images and GIS.
Environmental management, 2007Co-Authors: Guangjin Tian, Zhifeng Yang, Yaoqi ZhangAbstract:Through interpreting Landsat TM images, this study analyzes the spatial distribution of rural settlements in China in 2000. It calculates rural residential land percentage for every 1-km(2) cell. The entire country is divided into 33 regions to investigate the spatio-Temporal Dynamic patterns of rural residential land during the 1990s. According to the remote sensing survey, the rural residential land increased by 7.88 x 10(5) ha in the 1990s. The increment of rural residential land was 0.55 million ha in 1990-1995 and 0.23 million ha in 1995-2000. In 1990-1995, rural residential land increased dramatically in the eastern regions such as the Yangtze River Delta, Pearl River Delta, and North China Plain, accounting for 80.80% of the national growth; the expansion in the western regions was much more moderate. In 1995-2000, the expansion of rural residential land in eastern regions slowed, accounting for only 58.54% of the increase at the national level, whereas the expansion in the western regions accelerated. Rapid rural residential development resulted from increasing home construction and the limited control on rural land. The great regional disparity reflected the regional economic development and land-use policy change. Our finding shows that nearly 60% of the rural residential area came from cropland.
Abhinav Gupta - One of the best experts on this subject based on the ideXlab platform.
-
Temporal Dynamic Graph LSTM for Action-driven Video Object Detection
arXiv: Computer Vision and Pattern Recognition, 2017Co-Authors: Yuan Yuan, Xiaodan Liang, Xiaolong Wang, Dit-yan Yeung, Abhinav GuptaAbstract:In this paper, we investigate a weakly-supervised object detection framework. Most existing frameworks focus on using static images to learn object detectors. However, these detectors often fail to generalize to videos because of the existing domain shift. Therefore, we investigate learning these detectors directly from boring videos of daily activities. Instead of using bounding boxes, we explore the use of action descriptions as supervision since they are relatively easy to gather. A common issue, however, is that objects of interest that are not involved in human actions are often absent in global action descriptions known as "missing label". To tackle this problem, we propose a novel Temporal Dynamic graph Long Short-Term Memory network (TD-Graph LSTM). TD-Graph LSTM enables global Temporal reasoning by constructing a Dynamic graph that is based on Temporal correlations of object proposals and spans the entire video. The missing label issue for each individual frame can thus be significantly alleviated by transferring knowledge across correlated objects proposals in the whole video. Extensive evaluations on a large-scale daily-life action dataset (i.e., Charades) demonstrates the superiority of our proposed method. We also release object bounding-box annotations for more than 5,000 frames in Charades. We believe this annotated data can also benefit other research on video-based object recognition in the future.
-
ICCV - Temporal Dynamic Graph LSTM for Action-Driven Video Object Detection
2017 IEEE International Conference on Computer Vision (ICCV), 2017Co-Authors: Yuan Yuan, Xiaodan Liang, Xiaolong Wang, Dit-yan Yeung, Abhinav GuptaAbstract:In this paper, we investigate a weakly-supervised object detection framework. Most existing frameworks focus on using static images to learn object detectors. However, these detectors often fail to generalize to videos because of the existing domain shift. Therefore, we investigate learning these detectors directly from boring videos of daily activities. Instead of using bounding boxes, we explore the use of action descriptions as supervision since they are relatively easy to gather. A common issue, however, is that objects of interest that are not involved in human actions are often absent in global action descriptions known as “missing label”. To tackle this problem, we propose a novel Temporal Dynamic graph Long Short-Term Memory network (TDGraph LSTM). TD-Graph LSTM enables global Temporal reasoning by constructing a Dynamic graph that is based on Temporal correlations of object proposals and spans the entire video. The missing label issue for each individual frame can thus be significantly alleviated by transferring knowledge across correlated objects proposals in the whole video. Extensive evaluations on a large-scale daily-life action dataset (i.e., Charades) demonstrates the superiority of our proposed method. We also release object bounding-box annotations for more than 5,000 frames in Charades. We believe this annotated data can also benefit other research on video-based object recognition in the future.
Christopher K Wikle - One of the best experts on this subject based on the ideXlab platform.
-
Estimation of Parameterized Spatio-Temporal Dynamic Models
Journal of Statistical Planning and Inference, 2007Co-Authors: Christopher K WikleAbstract:Spatio-Temporal processes are often high-dimensional, exhibiting complicated variability across space and time. Traditional state-space model approaches to such processes in the presence of uncertain data have been shown to be useful. However, estimation of state-space models in this context is often problematic since parameter vectors and matrices are of high dimension and can have complicated dependence structures. We propose a spatio-Temporal Dynamic model formulation with parameter matrices restricted based on prior scientific knowledge and/or common spatial models. Estimation is carried out via the expectation-maximization (EM) algorithm or general EM algorithm. Several parameterization strategies are proposed and analytical or computational closed form EM update equations are derived for each. We apply the methodology to a model based on an advection-diffusion partial differential equation in a simulation study and also to a dimension-reduced model for a Palmer Drought Severity Index (PDSI) data set.
-
a kernel based spatio Temporal Dynamical model for nowcasting weather radar reflectivities
Journal of the American Statistical Association, 2005Co-Authors: Christopher K Wikle, Neil I FoxAbstract:A good short-period forecast of heavy rainfall is essential for many meteorological and hydrological applications. Traditional deterministic and stochastic nowcasting methodologies have been inadequate in their characterization of pixelwise rainfall reflectivity propagation, intensity, and uncertainty. The methodology presented herein uses an approach that efficiently parameterizes spatio-Temporal Dynamic models in terms of integro-difference equations within a hierarchical framework. The approach accounts for the uncertainty in the prediction and provides relevant distributional information concerning the nowcast. An application is presented that shows the effectiveness of the technique and its potential for nowcasting weather radar reflectivities.
