Periodicity Detection

Figure 1: Visualization of the matrix D of pairwise distances of the video frames, in which two periodic motions (jumping, waiving) appear after a non-periodic one (stand up). Warm (cold) colors correspond to high (low) values in D. Periodic segments are manifested as straight line segments in D (in white) that are parallel to the main diagonal and are associated with low sum of D values.

We present a solution to the problem of discovering all temporally periodic segments of a given video and of estimating their period in a completely unsupervised manner.

These periodic segments may be located anywhere in the original sequence, may differ in duration, speed, period and may represent unseen motion patterns of any type of objects (e.g., humans, animals, machines, etc).

The proposed method capitalizes on earlier research on the problem of detecting common actions in videos, also known as commonality detection or video co-segmentation [2-3].

A newer periodicity detection method is proposed in [4].

Experiments - Downloads

• You can download the matlab code (.rar) of PMUCOS method proposed in [1]
• You can download the videos, distance matrices and ground truth of PERTUBE (http://www.ics.forth.gr/cvrl/pd) dataset and the distance matrices and ground truth of MHAD202v dataset (.rar)
• You can download the videos of MHAD101v dataset (.rar) . MHAD202v is created by the MHAD101v by getting all video pairs.
• You can download the experimental results of PMUCOS presented in [1] (.rar).
• See the corresponding readme.txt files for more details.

Related Publications

[1] C. Panagiotakis, G. Karvounas and A. Argyros, Unsupervised Detection of Periodic Segments in Videos, ICIP, 2018.

[2] C. Panagiotakis, K. Papoutsakis and A. Argyros, A Graph-based Approach for Detecting Common Actions in Motion Capture Data and Videos, Submitted to Pattern Recognition, 2017.

[3] K. Papoutsakis, C. Panagiotakis and A.A. Argyros, "Temporal Action Co-Segmentation in 3D Motion Capture Data and Videos", In IEEE Computer Vision and Pattern Recognition (CVPR 2017), IEEE, Honolulu, Hawaii, USA, July 2017.

[4] C. Panagiotakis and A. Argyros, A two-stage approach for commonality-based temporal localization of periodic motions, International Conference on Computer Vision Systems, 2019.