Description:
UCF-CIL
Action Dataset consists of 56 sequences of 8
actions:
- 4 of ballet fouettes
- 12 of ballet spin
- 6 of push-up exercise
- 8 for
golf swing
- 4 of one-handed tennis backhand stroke
- 8 of two-handed tennis
backhand stroke
- 4 of tennis forehand stroke
- 10 of tennis serve
Each
action is performed by different subjects, and the
videos are taken by different unknown cameras from various viewpoints
collected
over Internet. In addition, videos in the same group (action) may have
different starting and ending times, thus may be only partially
overlapped.
Subjects also perform the same action in different ways and at
different
speeds.
We
provide point tracking for each action. The dataset is
arranged as follows: In the main directory, there are 8 different
folders for each action. Each action folder has a number of folders for
the
different examples. Inside each such folder, the image files
and the text files are provided. The image files are the different
frames of
the action and the corresponding text file contains the 11 points for
that
frame. The format of each text file is as follows: The first line gives
the
number of tracked points, which is always 11 in our case. This number
is
followed by 11 lines, each of which consist of 3 numbers. The first
number is
either 0 (denoting that the point is occluded) or 1 (point is
visible). The next two numbers
are the x and y coordinate of the point. The 11 points are always in
this
order: head, right shoulder, right elbow, right hand, left
shoulder, left elbow, left hand, right knee, right foot, left knee, and
left
foot.
|
Publications:
- Yuping Shen and Hassan Foroosh, View Invariant
Action Recognition from Point Triplets, IEEE Transactions on Pattern
Analysis and Machine Intelligence (PAMI), to appear, 2009. (PDF)
- Yuping Shen and Hassan Foroosh, View Invariant
Action Recognition Using Fundamental Ratios, Proc. IEEE Conference on
Computer Vision and Pattern Recognition (CVPR), 2008. (PDF)
- Yuping
Shen
and Hassan Foroosh, View Invariant Recognition of Body Pose from
Space-Time Templates, Proc. IEEE Conference on Computer Vision and
Pattern Recognition (CVPR), 2008. (PDF)
|