In recent times, analytics has created a niche for itself in the sports industry. With increasing competitiveness, management, teams and athletes have realized the need to complement their training with machine learning and artificial intelligence-led video analytics. With the ability to host and analyze in cloud, AI-led video analytics can be leveraged to enable greater efficiencies.
This article demonstrates a framework for identifying and using analytics on certain key performance indicators and metrics that can help athletes understand and optimize their performance. This framework can be applied to a wide variety of sports and athletes. With the use of video analytics, athletes will be able to measure themselves against competition and find areas of improvement, if any. The framework can be customized to a particular sport and key metrics can be defined accordingly.
Framework for AI-led video analytics in sports
The framework emphasizes on three different areas - recording the video during the event, followed by competition analysis, and finally, training analysis.
From data collection point of view, it will provide both options of near real-time analysis and post-match analysis. From an analysis point of view, it will provide both, competition and training analysis. In competition analysis, it will concentrate on post-match analysis like generating performance profile of every athlete, creating a comparison among players based on different parameters, and analyzing how these affected the overall result of the match. In training analysis, it will concentrate on checking the progress of athletes by creating custom drill sessions to improve the athletes’ weaknesses to achieve optimal performance.
How does this framework work?
To begin with, we will collect data on performance of athletes using video recordings during the live sport event. The recorded videos will be uploaded into the video analytics application. For this process of data collection, on the hardware front, we will need access to the video recordings by professional camerapersons or we can use any portable cameras. The cameras need to be calibrated appropriately in accurate angles to cover the movement of athletes in the field. The data collected then will be uploaded real time or after the match to the cloud for processing via mobile application or website.
The next step involves analysis of the data captured. To begin with, we will use facial recognition to identify the athletes. The analysis can be performed on near real time or after the event. Necessary key performance metrics will be extracted. Goals and objectives will be set and the performance will be measured against these. Coaches and athletes together can decide on desired goals according to their needs. These KPIs will vary with different sports. Some of the common metrics include overall average athlete speed, distance and trajectory. Some of the sports-specific KPIs include number of passes made by a player, basket count, accuracy in case of football and basketball.
The concluding step is to provide analysis in the form of interactive dashboards that deliver real time feedback and help the athlete to improve his/her activity. The dashboard will be available via the mobile app on the device. The athlete will be capable of continuous improvement via accurate tracking of progress.
The way forward
Some of the key challenges in enabling AI-led video analytics for sportspersons include appropriate infrastructure availability i.e. video cameras and related accessories, servers to run the algorithms to analyze the performance data if cloud is not preferred, the accuracy of algorithms to provide appropriate information, and good connectivity for real time analysis.
The inferences drawn with analytics can also keep the sports fan engaged and connected. Fans around the world will be able to follow their favourite teams and athletes and virtually be part of their journey. The growth of smart wearables market will greatly complement the analysis using video analytics in the foreseeable future.