Content
There are two types of sequences:
real-lecture videos and challenges videos. The real-lecture category consists
of two long duration videos (30 minutes each) that have been recorded in a real
lecture; they present some difficulties as target disappearing and reappearing
on the scene, target appearance changes and scene illumination changes.
In addition to the two long videos, 10
short different challenges videos designed to include difficult situations that
a tracking algorithm might face during a lecture have been recorded with the
aim of testing the algorithm in these situations. C1 presents the target taking
off a sweater, which represents a high appearance change. In C2 and C3, the
tracked target leaves the scene and then comes back to it. In C2, the exit and
entry are produced in the same place, different than C3 in which they are produced
at different locations. A significant occlusion is shown in C4, where the
target moves and hides behind the blackboard coming out on the opposite end of
it. In C5, the target squats (for example, to turn on a computer) behind a
table. The challenge presented in C6 is similar than the one presented in C4,
but in this case the target comes in on the same place where he came out. Both
C7 and C8 present people and target crossing situations with people wearing
similar (dark) clothes. C9 shows a situation where the target sits down and
waits while two people with similar appearance go on stage sequentially.
Finally C10 shows a target spinning.
|