Professor Jie Ding

Data in the form of streaming videos are common in everyday life. These data are usually massive and require much storage and communication bandwidth. Inference based on the videos such as anomaly detection and semantic analysis are even more computational challenging.

The goal of this project is to develop new architectures for compressing videos and performing statistical inference based on the compressed videos.  The project will focus on deep neural net based architectures and develop algorithms such as change point analysis and semantic labeling based on the compression domain. Therefore, the group's efforts heavily rely on GPU computing that parallelizes the training and testing of algorithms.