Automatic Activity Detection in video-sequences
Event-Based indexing consist in automatically detecting key
moments corresponding to a particular activity into long video sequences. The
techniques developed for indexing can be used to detect the activities done
by a subject in a given video-sequence.
To this aim, Manor and Irani [1] developed a similarity statistical measure
based on spatio-temporal features. This measure can be used for isolating and
clustering events without any prior knowledge of the types of events, their
models, or their temporal extent.
Figure 1: Activity segmentation from video-sequences [1]
In this project the student should implement such clustering
measure and test the performances in several video-sequences that he/she will
capture using the HW (cameras) available in the lab.
Reading:
[1] L.Zelnik-Manor and M. Irani, "Event-Based Video Analysis", IEEE Conference on Computer Vision and Pattern Recognition (CVPR), December 2001