School of
Information Technology and Electrical Engineering

Speaker: Mr Chao Li
Seminar Date: Wed, 29/11/2017 - 13:30
Venue: 78-631; Data Science MMLab
Host: Dr Helen Huang

Seminar Type:  PhD Thesis Review


Nowadays, video is becoming the most popular means of documenting everything from recipes to how to change a tire of a car. Technologies for automatically understanding video events are in high demand for analysing and managing the exploding video content.

In this thesis, we shows how to discover temporal concepts/patterns and combine them with static visual features for video event classification and retrieval, which are two main tasks about video content understanding. This thesis consists of four main chapters. In the first chapter, we propose a data-driven hierarchical structure of latent variables to discover latent concepts for event classification. In the second chapter, we incorporate weak semantic relevance, as fine-grained guidance (at shot-level) to the proposed temporal attention model, to facilitate video event classification. In the third chapter, we jointly model two essential aspects of videos, i.e., temporal pattern and static visual feature, for unsupervised video event retrieval. In the fourth chapter, we propose a novel information filtering mechanism named as Adaptive Selection, which exploits the complementary advantages from the two above aspects for supervised video event retrieval.