Open Research Challenge: Video analysis / statistics

In order to evaluate the progress of the league, regardless of the annual rule changes, we need statistics similar to human soccer.

Therefore, only GameController/TeamCom statistics are not sufficient for this.

That’s why there is an Open Research Challenge, as already known from the year 2019 (and previously until RoboCup 2014), which will focus on the generation of statistics from external video data from a go-pro type camera viewpoint.

Challenge Goal

For this Open Research Challenge two major goals exist.

The more short-term goal is to calculate extrinsic camera parameters (camera matrix) from the camera feed and to locate/track all moving objects (ball, robots) on the field.

In addition to this, the more long-term goal involves the creation of game statistics based on the located objects and positions from the short-term goal.

Here, the following statistics, among others, might be of interest, such as time under control, shots on goal (successful/unsuccessful), passes and so on.

Since this is an open challenge, the decision of what to show (short-term, long-term, parts of it) is, within the scope of the goals aforementioned here, entirely up to the teams.

In addition, the execution time for this challenge is also not relevant for the time being whether you process it online or offline to the video stream.