Identifying Surprising Events in Video, and Foreground/Background Segregation in Still Images

When?

Friday 21 September 2012, 15:00 to 16:00

Where?

CVSSP Seminar Room (40bAB05)

Open to:

Public, Staff, Students

Speaker:

Prof Daphna Weinshall, School of Computer Science and Engineering, Hebrew University of Jerusalem

Abstract:

Automatic processing of video data is essential in order to allow for efficient access to large amounts of video content, a crucial point in such applications as video mining and surveillance. To this end, in the first part of the talk I will describe a method which identifies for the human viewer interesting parts of the video. The method employs the notion of Bayesian surprise, in which an event is considered surprising if its occurrence leads to a large change in the probability of the world model. We specifically measure the change in the Dirichlet prior of the LDA world model as a result of each video event's occurrence. We tested our algorithm on a real dataset of video data, taken by a camera observing an urban street intersection. The results demonstrate our ability to detect atypical events, such as a car making a U-turn or a person crossing an intersection diagonally.

A related problem with large collections of still images would seek to segregate foreground from background, in order to identify the location of objects on interest in the image. Effective segmentation prior to recognition has been shown to improve recognition performance. Accordingly, in the second part of the talk I will describe a method for the extraction of foreground masks. Our method segments the interesting objects from the rest of the image without any class specific knowledge of object categories; it uses a probabilistic formulation which is derived from manually segmented images, which is trained on the fly from images that are similar to the query image. The method was tested on the challenging VOC09 and VOC10 segmentation datasets, achieving excellent results in providing foreground masks.