Seminar: Exploiting Geospatial Properties for Efficient Visual Data Management and Learning

A PDF of the presentation is available here: CS Seminar 8 - Presentation Materials.

Speaker

Seon Ho Kim, PhD

Research Scientist and Associate Director at the Integrated Media Systems Center, USC

Abstract

Due to the ubiquitous availability of smart mobile cameras such as smartphones and drones, a number of trends have recently emerged with respect to visual data management. First, we are experiencing unprecedented growth in the amount of visual content that is being collected from mobile devices. Creating, searching, sharing and viewing images and videos are immensely popular in lots of applications. Second, mobile cameras include a plethora of built-in sensors such as global position system (GPS) receiver and digital compass, recording geospatial properties (e.g., location and viewing direction of camera) during capture time. This facilitates the modeling of mobile visual content through its geo-spatial properties at a fine granular level and provides an essential metadata to manage a large amount of geo-tagged visual data. This talk focuses on how to utilize geospatial metadata of images and videos for efficient data management and learning since location information is essential in many database applications and provides a fundamental connection in managing and sharing visual data among users. It describes how such geospatial metadata enhance the collection, searching, sharing of visual data, especially at scale. Furthermore, this talk addresses how to harness spatial metadata for efficient image-based machine learning, not only for a highly efficient visual data management, to prepare for upcoming era of machine learning and AI. Real world use cases with real datasets will be introduced for a better understanding.

Bio

Dr. Seon Ho Kim is a Research Scientist and Associate Director in the Integrated Media Systems Center (IMSC) at the University of Southern California. He also teaches in the Data Science Program in the Viterbi School of Engineering at USC.

He received his BS degree in Electronic Engineering from the Yonsei University, Seoul, Korea in 1986. He received his M.S. in Electrical Engineering and Ph.D. in Computer Science from the University of Southern California in 1994 and 1999 respectively. Between 1999 and 2010, he worked as a faculty member at University of Denver, Colorado, and University of the District of Columbia, Washington DC.

Dr. Kim's primary research interests include multimedia systems, databases, Big Data, image machine learning, IoT, and mobile media data management, where he has more than 90 publications including a textbook, patents, major journal papers, and top conference papers. He has been leading MediaQ project which is a social media data collection and management platform, and recently TVDP project which focuses on a translational visual data platform for urban geo-tagged visual data collection, processing, and analysis. Recently, his work received the best paper awards at the IEEE Big Multimedia Conference two years in a row in 2018 and 2019. His team also received the 1st Place award at the IEEE Big Data Challenge Cup in 2020 (RDD: Global Road Damage Detection). Dr. Kim’s projects have been funded by US federal and local government, industries such as Google, Northrop Grumman Corp., Microsoft, Oracle, and Intel.

He worked as the president of KSEA-SC (Korean American Scientists and Engineers Association Southern California Chapter) and the president of KOCSEA (Korean Computer Scientists and Engineers in America). Dr. Kim is also the founder and president of SoDAVi (Social Data Analysis and Visulization) Group, a non-profit organization which aims for providing data collection and analysis service to Korean-American community in Southern California.

Meeting details will be emailed to the faculty and students.

Visit the CS Seminars Page for more information about Computer Science Seminars