Could you imagine an AI observer in esports? It might be closer than you think cover image

Could you imagine an AI observer in esports? It might be closer than you think

Intelligent automatic observers might be the future of the esports viewing experience. Read more about the study here.

If you have consumed enough esports content, you might notice that some games are difficult to watch due to unsteady or wonky observing. Either the camera failed to capture big moments, it was too hectic, or too slow - these can lead to a bad viewing experience.

A group of researchers from Gwangju Institute of Science and Technology (GIST) in South Korea attempted to solve the issue. They have created an intelligent, automatic observer for esports using a new technology that can mimic great human observing work.

Of course, there are already existing automatic observers that can do a pretty fair job. But the new technology, created by these researchers, sees better results in terms of finding scenes of interest for the audience.

Explaining how the new technology works

The study involves a new object detection algorithm, Mask R-CNN which uses a different approach to detect objects. While the conventional way detects a single unit - for example, a worker or a building - as an object, the Mask R-CNN treats a 2D spatial area viewed by the spectator as an object.

The study used 25 StarCraft games to collect data.<br>Image via Interface In Game
The study used 25 StarCraft games to collect data.
Image via Interface In Game

Next, they collected 25 human-observed games from the real-time strategy title, StarCraft. They used Mask R-CNN to detect the objects, which is the area viewed by the spectator. This means that the system had marked interesting events by detecting areas instead of units. But they also use in-game features as input data.

Using another system, they identified patterns to find the “region of common interest” (ROCI). The ROCI is the most exciting area for the spectators to watch.

This results in a framework that is more successful than those using pre-existing methods. The automatic esports observing work sees a higher similarity to human observational data. This method also sees a better performance in more unique games with different starting locations and maps. 

According to Dr. Kyung-Jong Kim, Associate Professor at GIST, this newly-founded system can also be applied to other games. “The framework can be applied to other games representing some of the overall game state, not only StarCraft,” he stated.

He also talked about developing this new technology for future uses. “As services such as multi-screen transmission continue to grow in Esports, the proposed automatic observer will play a role in these deliverables. It will also be actively used in additional content developed in the future.”

You can find the full study here.

A positive change for esports observing

Image via Valve
Image via Valve

The role of a game observer is indeed vital for a smooth viewing experience in esports. In this age where esports has grown into a gigantic industry, having a reliable automatic observer would be very efficient. 

With this new technology, there would be lesser human errors that can possibly ruin a spectator’s esports experience. Hence, the future of esports could see a massive improvement in terms of quality.

Stick around for more esports news and updates!