In the growing field ofe-sportresearchers are looking into artificial intelligence to better enjoy the show.
More specifically, the Institute of Science and Technology of Gwangju (ISTG) in South Korea, a veritable eldorado of this discipline, is working to develop an intelligent observer so as not to miss any major action.
Automatic observer search to ensure the show
On flagship e-sport games like counter strike Where League of Legends, the follow-up of a match is ensured by human observers controlling the camera, and the commentators provide their analysis in relation to the action displayed on the screen. But with professional players scattered all over the map, there may be times when decisive action escapes their vigilance.
This is why the demand for automatic observers has recently experienced a certain boom. However, it is still difficult at present to develop such a feature capable of capturing the best moments of a game.
This may change with research conducted by the ISTG. ” We created an automatic observer using an object detection algorithm to learn data from human spectators “, thus indicates in his study Dr Kyung-Jong Kim, at the head of the project.
Follow the eyes of the spectators
The idea is therefore to define the object to be observed as the two-dimensional space seen by the spectators, and not as a single object to be observed, which is what object detection currently deals with. As a point of reference, the researchers collected sighting data from 25 spectators on part of StarCraft 2another major esports title in South Korea, and highly complex to follow for human observers.
The researchers then applied a distinction between the areas observed by the spectators, identified by the unit “one”. The rest of the screen is thus filled with zeros, so that the automatic observer is only interested in areas that catch the eye of the spectators.
All of this data fed into the automatic observer’s neural network to effectively define “regions of common interest”. In their preliminary tests, the researchers noticed that their creation was much more effective than the automatic observers developed so far.
Dr Kyung-Jong Kim believes that this project could bring many benefits for spectators of e-sports events, in particular with the development of multi-screen transmissions. The other side of the coin is that human observers, whose services are sometimes considered too expensive for low-budget competitions, could soon find themselves on the sidelines.
Source : ScienceDirect
2