Analysis of open tracking data, 02, several plottings

On the paper and data, this page。

"Soccer video and player position dataset": S. A. Pettersen, D. Johansen, H. Johansen, V. Berg-Johansen, V. R. Gaddam, A. Mortensen, R. Langseth, C. Griwodz, H. K. Stensland, and P. Halvorsen, in Proceedings of the International Conference on Multimedia Systems (MMSys), Singapore, March 2014, pp. 18-23 

Plottings

First, tracking plot of ID# 8. The left defensive midfielder?

Units of the x- and y-axis are [m].

As is shown in the above figure, overlapped is the line showing the player's track. So, one cannot found where is the frequent position. `Heat map' is useful in this case.

The field is divided into the unit area (dx = 5.2 m, and dy = 5.7 m). The color bar indicates the integrated time in each area.

In this case, the longest integrated time is 113.5 seconds. Excluding an area of 0 second (black area), the median value is 19.4 seconds.

From this heatmap, one can guess that this player is the left defensive midfielder.


Next, the following image shows temporal variations of the velocity (top) and of the direction (bottom), from 40 min to 45 min.

The dashed line in the top panel indicates the maximum velocity in the 1st half, 9.3 m/s.

(The other players show the maximum velocities like this. It is not fast. Why?)

In the bottom panel, the positive direction of the y-axis indicates 0 radian.

The data range is from -\pi to \pi. So -\pi/2 and \pi/2 are goal directions (the dashed lines).

 But it is not clear which goal direction is positive value.

Temporal variations shown in these panels have time scales around 1 second.


Activity Time Spent shown in 2018 FIFA World Cup Russia is estimated from the velocity variation like this.

Data quality

The data sizes of Petterson et al. are ~80MB.

It is easily analyzed with modern PC performances.

But some data are lost. And tracking data of balls are not enough.

Now I checked only the 1st half data of 2013-11-03, and found that lost are data of #3, #6, #11, and #12.

We should check other data sets.

One candidate is The DEBS 2013 Grand Challenge.

But the compressed data size is 2.6GB. It is too large!

One needs a PC having a large memory. Consumer products are not enough.

Memory size of `colab' is 13 GB. This would be not enough.



Notebook for the tracking plot and the heatmap.

Notebook for the velocity panels.


[articles] Tracking Data