Long-Term Recurrent Convolutional Networks for Visual Recognition and Description

Multitemporal environmental and urban studies are essential to guide policy making to ultimately improve human wellbeing in the Global South. Land-cover products derived from historical aerial orthomosaics acquired decades ago can provide important evidence to inform long-term studies. To reduce the manual labelling effort by human experts and to scale to large, meaningful regions, we investigate in this study how domain adaptation techniques and deep learning can help to efficiently map land cover in Central Africa. We propose and evaluate a methodology that is based on unsupervised adaptation to reduce the cost of generating reference data for several cities and across different dates. We present the first application of domain adaptation based on fully convolutional networks for semantic segmentation of a dataset of historical panchromatic orthomosaics for land-cover generation for two focus cities Goma-Gisenyi and Bukavu. Our experimental evaluation shows that the domain adaptation methods can reach an overall accuracy between 60% and 70% for different regions. If we add a small amount of labelled data from the target domain, too, further performance gains can be achieved.

Download Full-text

Pruning Long-term Recurrent Convolutional Networks for Video Classification and captioning

2020 International Conference on Smart Electronics and Communication (ICOSEC) ◽

10.1109/icosec49089.2020.9215414 ◽

2020 ◽

Author(s):

Saikiran Gogineni ◽

G. Suryanarayana ◽

K.L.S. Soujanya

Keyword(s):

Video Classification ◽

Convolutional Networks

Download Full-text

Hybrid Spatial–Temporal Graph Convolutional Networks for On-Street Parking Availability Prediction

Remote Sensing ◽

10.3390/rs13163338 ◽

2021 ◽

Vol 13 (16) ◽

pp. 3338

Author(s):

Xiao Xiao ◽

Zhiling Jin ◽

Yilong Hui ◽

Yueshen Xu ◽

Wei Shao

Keyword(s):

Smart Cities ◽

Convolutional Networks ◽

Temporal Correlations ◽

Spatial Features ◽

Temporal Features ◽

Prediction Scheme ◽

Temporal Graph ◽

Real World Datasets ◽

The Internet Of Things

With the development of sensors and of the Internet of Things (IoT), smart cities can provide people with a variety of information for a more convenient life. Effective on-street parking availability prediction can improve parking efficiency and, at times, alleviate city congestion. Conventional methods of parking availability prediction often do not consider the spatial–temporal features of parking duration distributions. To this end, we propose a parking space prediction scheme called the hybrid spatial–temporal graph convolution networks (HST-GCNs). We use graph convolutional networks and gated linear units (GLUs) with a 1D convolutional neural network to obtain the spatial features and the temporal features, respectively. Then, we construct a spatial–temporal convolutional block to obtain the instantaneous spatial–temporal correlations. Based on the similarity of the parking duration distributions, we propose an attention mechanism called distAtt to measure the similarity of parking duration distributions. Through the distAtt mechanism, we add the long-term spatial–temporal correlations to our spatial–temporal convolutional block, and thus, we can capture complex hybrid spatial–temporal correlations to achieve a higher accuracy of parking availability prediction. Based on real-world datasets, we compare the proposed scheme with the benchmark models. The experimental results show that the proposed scheme has the best performance in predicting the parking occupancy rate.

Download Full-text

A new opportunity for the emerging tellurium semiconductor: resistive switching device implementation

10.21203/rs.3.rs-273013/v1 ◽

2021 ◽

Author(s):

Yifei Yang ◽

Mingkun Xu ◽

Lujie Xu ◽

Xinxin Wang ◽

Huan Liu ◽

...

Keyword(s):

Data Storage ◽

Resistive Switching ◽

High Performance ◽

Visual Recognition ◽

Stimulus Frequency ◽

Dielectric Materials ◽

Cross Point ◽

Electrochemically Active ◽

Event Based

Abstract The electrochemical (EC) resistive switching (RS) cross-point arrays, composed of nonvolatile RS (NV-RS) memories and volatile RS (V-RS) selectors, hold promise for high-density data storage, in-memory computing and neuromorphic computing. However, the conventional EC-RS devices based on metallic filaments suffer from the notorious current-volatility dilemma that the low and high current requirements for NV-RS memories and V-RS selectors, respectively, cannot be satisfied simultaneously, due to the dominant EC nature of the RS. In this work, we demonstrate electrochemically active, low thermal-conductivity and low melting-temperature semiconducting tellurium filament-based RS devices that solve this dilemma, enabling NV-RS memories to operate under lower currents than do V-RS selectors. This novel phenomenon arises as the consequence of the adversarial EC and Joule heating (JH) effects. The devices also show unusual stimulus frequency dependent long-term plasticity (LTP)-to-short-term plasticity (STP) transition. Devices with this property can be generically utilized as spatial-temporal filters in spiking neural networks (SNNs) for high-performance event-based visual recognition tasks, as illustrated in our noise filtering simulations. By regulating the EC-JH relationship using dielectric materials with decreasing thermal conductivities, full functional-range tunable Te filament-based devices, from always-NV RS, to NV-to-V transitionable RS, and to always-V RS, are also demonstrated. The tellurium filament-based RS devices are promising enablers for functional cross-point arrays.

Download Full-text