cSpider: Cloud-based System for Multimedia Content Protection: We designed a cloud-based video copy detection system for large-scale 3D video collections, called cSpider. The system creates compact signatures from videos that are fast to compute and compare, small to store, and robust against video modifications. The system maintains a large-scale distributed index of signatures on a cloud infrastructure. Utilizing the cloud, the system can scale (up and down) to different number of videos.
cSpider is currently running on the Amazon cloud and on our own local cloud. See a demo of cSpider
Distributed and Approximate Big Data Analysis: This project aims to enable important machine learning algorithms to efficiently process massive datasets. We achieve this goal using two techniques: controlled approximation and elastic distribution of computation.
gCloud: System for On-demand Sharing of GPUs in Clouds: This project develops novel methods to efficiently share GPUs among different applications in fine-grain and on-demand manner, yielding high utilization of GPUs and reduction in the execution times of applications. The developed methods and algorithms have been integrated in the popular OpenStack cloud management software system, which is backed up and used by many companies worldwide.