Ciklum Data & Analytics team member won The Cybersport Challenge Hackathon, aimed at finding technological solutions for the esports industry. Oleg Panichev, a Senior Research Engineer at Ciklum, and his team “Nullset” developed the computer vision algorithm to perform automated advertising analytics for cybersport broadcasting.
Gaming has long stopped being solely an entertainment for the solitary player holed up in their living room. In 2019, there were 201.2 million esports enthusiasts in the audience, along with 252.6 million occasional viewers. The entire eSports market is expected to grow over the coming years, with worldwide revenues projected to reach 1.79 billion U.S. dollars in 2022. During 2020, the merchandise and tickets segment in the global eSports market was expected to generate 106.5 million U.S. dollars in revenue, making this a huge playing field for the content creators and advertisers.
Esports content production companies and advertisers place virtual billboards, signage and sponsored characters in the games as well as sponsored experiences. After the gaming events, the advertisers and publishers report to brands on the activities and ROI. Usually, the processing of the brand appearances and ads impressions are done manually, which is not only a time-consuming job, but it’s also missing a part of important analytics and stats which are not possible to calculate manually. In addition, this job is usually done by different providers – there are providers who recognise brand logos on the video, and separately those who make advertising analytics.
The Nullset team has created a system that combines the latest Computer Vision technologies and automatically recognises brand logos on the tournament video. Based on this information, it automatically provides detailed analytics in a convenient format.
Demo on the example of the Na’Vi team game streaming during the hackathon
The demo shows how the system detects logos highlighted in green. After that, it automatically makes the report with calculated metrics, such as:
- the number of times the logo appears in the video
- how long the logo is displayed
- % of the video area
- % of a particular logo display in relation to others
- % of placement: banner ad or player’s t-shirt
- % of placement: in-game/not in-game
The Nullset team won the first prize with $ 3,000 award in the Cybersport Challenge Hackathon among 35+ teams. Having positive feedback from the companies during the event, the team have started working on scaling the project for future activities:
- The client uploads the video on the website or sends a link.
- Video is analysed on a powerful server, which allows processing large volumes of data quickly.
- Computer Vision algorithm recognises the complex logo placements, additional metrics and financial performance of the forecasting models.
- The client receives a report in a convenient format.
The cybersports industry is now surviving exponential increase in viewers and brands attention. Brands are looking for a positive way to interact with a passionate audience, and we are glad that our model aimed to make the life of publishers and advertisers easier was recognised. We believe that this solution will bring significant results for the companies.Oleg Panichev, Senior Research Engineer at Ciklum
Looking for a digital partner to implement your deep learning project? Contact Ciklum and we’ll get back to you within 24 hours!