Live Content Processing System for Mutual Society Safety Interactions
Prof. Aran Hansuebsai
|Department of Photographic Science and Printing Technology|
Dr. Charasroj Bothdamrih
Telecommunications nowadays can quickly reach a large group of people anywhere, any places in real time, using social media applications mainly for commercial or entertainment purposes. Adversely, natural disasters and terrorist attacks are occurring more frequently worldwide so people should be able to receive news and updates about these potential signals of serious risks for safety information. Though, information about natural disasters and terrorist attacks are currently used limitedly and under the control of government agencies.
The purpose of the study is to examine and test the connectivity of live sensors from CCTVs at various locations and display multiple video streams simultaneously on the screen. As a result, the general public will be able to view these videos online in real time via cellular network and the Internet. The result of final data processing and communications shows that it takes 3 – 5 seconds to reach the end user within optimum number of hops in normal network environment. We have found that this crucial application needs strong support and collaboration from certain organizations. These organizations include the ones who own different types of sensors to be combined and the Internet service providers who can allow this service to happen by connecting data systems and managing the communications to people's mobile devices. So, people in different locations will have access to the information from their mobile device and be informed about the warnings as soon as any incident occurs. At the same time, people would be able to send live feeds videos to this platform. We also aim to actuate the collaboration of all agencies both public and private sectors to overcome the constraints and attempts to design a new platform to be the Mutual Social Safety Interactions with clear and provable validations.
Dimension Reduction for Visualizing High Dimensional Data
|Prof. Pitoyo Hartono|
Department of Electrical and Electronic Engineering
School of Engineering
The proliferation of Big Data increases the importance for dimensional reduction techniques that are essential for visualizing high dimensional data. Visualization offers intuitive means for understanding the underlying properties of high dimensional data and thus applicable over wide range of fields. In this talk, I will start by reviewing the theoretical aspects and some examples of traditional dimensional reduction and visualization techniques such as Principal Component Analysis (PCA), Multidimensional Scaling (MDS) and Self-Organizing Maps (SOM) and proceed into more recent approaches such as Restricted Boltzmann Machine (RBM), which is an essential part of Deep Belief Networks, and Stochastic Neighbor Embedding (SNE). I will also explain about the theory and applications of my current study of Context-Relevant Self-Organizing Map.