Research‎ > ‎

Project: Video Visual Analytics

Interactive Video Visual Analytic (Human Centred Computer Vision)

Related Software: sViSIT

Related Publications: 

Amir H. Meghdadi and Pourang Irani, "Interactive exploration of surveillance video through action shot summarization and trajectory visualization", the IEEE Conference on Visual Analytics Science and Technology, 2013.  

The goal of this research is designing interactive systems that use both computer vision algorithms and the human user in close interaction with each other to achieve a goal or a task. This is important in applications where the task is either too complicated for machine intelligence or it is subjective and context dependent. For example, analysis of surveillance videos in search of important events is a challenging task for computer vision because often the criteria for detecting the importance of the events is vague, context dependent and subjective. The task is also challenging for a human user because these videos are very long and browsing through the footage is very tedious and time consuming. Our solution to this problem is interactive video summarization where computer vision is used to detect, summarize and visualize movement events as the human user interacts with the system to query the content and make decisions about the importance and the nature of events. We  developed a system (sViSIT) for selective and interactive visualization and summarization of the events in videos. This system uses both the computational power of computer vision algorithms and the intelligence and judgement of a human user in a closed loop interaction. Figures below show sample output images of the system that summarize two sample events in two different videos. 

Computandia Amir H. Meghdadi,
Oct 4, 2013, 11:33 AM