Social and Spatio-Temporal Learning for Contextualized Next Points-of-Interest Prediction

Author(s):  
Sayda Elmi ◽  
Karim Benouaret ◽  
Kian-Lee Tan
Author(s):  
Ching-Hang Chen ◽  
Tyng-Luh Liu ◽  
Yu-Shuen Wang ◽  
Hung-Kuo Chu ◽  
Nick C. Tang ◽  
...  

2021 ◽  
pp. 447-456
Author(s):  
Matthias Seibold ◽  
Armando Hoch ◽  
Daniel Suter ◽  
Mazda Farshad ◽  
Patrick O. Zingg ◽  
...  

2020 ◽  
Vol 153 ◽  
pp. 02003
Author(s):  
Putu Edi Yastika ◽  
Norikazu Shimizu ◽  
Ni Nyoman Pujianiki ◽  
I Gede Rai Maya Temaja ◽  
I Nyoman Gede Antara ◽  
...  

Numerous cities around the world are facing the problem of land subsidence. In many cases, it is the excessive groundwater extraction to meet human needs that leads to this subsidence. Since land subsidence rates are very slow (a few centimeters per year), the subsidence usually remains unnoticed until it has progressed to the point of causing severe damage to buildings, houses, and/or other infrastructures. Therefore, it is very important to detect the presence of subsidence in advance. In this study, screening for the presence of land subsidence in the city of Denpasar, Bali, Indonesia is conducted. The Sentinel-1A/B SAR dataset, taken from October 2014 to June 2019, is processed using the SBAS DInSAR method. Subsidence is found in the districts of Denpasar Selatan, Denpasar Barat, and Kuta, which falls in the range of -100 mm to -200 mm in an area of about 93.03 ha. All the extracted points of interest show the subsidence having linear behavior. The spatio-temporal behavior of the subsidence in Denpasar is presented clearly. However, the mechanism and the deriving factors of the subsidence remain unclear. Therefore, further studies are needed.


2019 ◽  
Vol 28 (1) ◽  
pp. 291-301 ◽  
Author(s):  
Kaihao Zhang ◽  
Wenhan Luo ◽  
Yiran Zhong ◽  
Lin Ma ◽  
Wei Liu ◽  
...  

2020 ◽  
Vol 10 (2) ◽  
pp. 557 ◽  
Author(s):  
Mei Chee Leong ◽  
Dilip K. Prasad ◽  
Yong Tsui Lee ◽  
Feng Lin

This paper introduces a fusion convolutional architecture for efficient learning of spatio-temporal features in video action recognition. Unlike 2D convolutional neural networks (CNNs), 3D CNNs can be applied directly on consecutive frames to extract spatio-temporal features. The aim of this work is to fuse the convolution layers from 2D and 3D CNNs to allow temporal encoding with fewer parameters than 3D CNNs. We adopt transfer learning from pre-trained 2D CNNs for spatial extraction, followed by temporal encoding, before connecting to 3D convolution layers at the top of the architecture. We construct our fusion architecture, semi-CNN, based on three popular models: VGG-16, ResNets and DenseNets, and compare the performance with their corresponding 3D models. Our empirical results evaluated on the action recognition dataset UCF-101 demonstrate that our fusion of 1D, 2D and 3D convolutions outperforms its 3D model of the same depth, with fewer parameters and reduces overfitting. Our semi-CNN architecture achieved an average of 16–30% boost in the top-1 accuracy when evaluated on an input video of 16 frames.


Sign in / Sign up

Export Citation Format

Share Document