In a country where people worship cricket as a religion, Qudich Tracker aims to change how matches are viewed on television through an innovative approach that uniquely bridges the gaps in storytelling around fielding and fielding strategies.

As cricket fans, there is an innate desire to know what is happening on the pitch during critical moments of the match. Fielding positions are changed, and a lot of drama occurs near the boundary lines. But the camera cannot show us all these events in one frame. As a result, fans are primarily deprived of all the storytelling around fielding strategy. 

The existing setup of viewing a cricket match offers no visual cues about the field placements and fielding changes. Broadcasters had no reliable visual to represent fielding strategy and changes. As a result, viewers are clueless about the rationale for bowlers to bowl a particular line or length. They also do not get the batsman's viewpoint when they choose to play a shot in a specific direction. This accounts for a relatively poor viewing experience. In addition, the teams and coaches do not have any fielding data and manually log basic information. 

The only possible way to achieve this in real-time with a single camera was to train a Computer Vision algorithm based on real data from a cricket field. However, any analytics-based solution would not have delivered the flexibility and accuracy required for a live broadcast. 

The core AI behind QT is a computer vision-based deep learning algorithm trained on over 750k curated images. In addition, it provides a highly accurate detection and tracking module that tracks every player on the field 25 times in a second. 

It also generates over 5 million data points per T20 game which can be used for further analysis. All the analytics and visualization are performed in real-time at 25 FPS with only a 4-frame delay, less than 1/5th of a second.

This AI-based solution is best suited for the problem as it solves multiple issues like details like variations in conditions, including light changes, team jersey colours, and color of the grass, which an analytics-based approach cannot capture. Additionally, the overall accuracy and the requirement of being real-time in a live scenario make any analytical solution redundant.

What makes QT unique?

The Quidich Tracker tracks all the players' live positions and movements on a cricket field. The television viewers can visualize the real-time field plot of the entire cricket ground in the form of a 3D digital replica of the stadium with dots representing the players during the live broadcast. 

The architecture of the tracker is designed in a manner where 4k video input is taken via a capture card. A capture card is a hardware input device that converts video signals to digital data that can be uploaded to the internet. The pixel coordinates are then captured by a detection and tracking module, which then gets converted into real-world coordinates. This finally gives the visualization of a 3D model of the stadium. 

Highly accurate detection, heuristics-based approach to managing occlusions, single-camera solution, and minimal requirement of physical resources make Quicich Tracker a unique solution. Unlike manual plot, QT is always-live available at the director's command whenever field changes occur . It also offers real-time distances, gaps etc., to enhance storytelling and viewers' understanding of fielding strategy.

Its Impact

 In the past few years, there has been high adoption of QT. As a result, it is now integral to every significant cricket broadcast worldwide (ICC World Cups, IPL, CPL, and Bilateral Cricket Series). On its debut at ICC World Test Championships Final, QT was used more than 200 times during the match. It also substantially helped increase discussions and analysis by commentators on live fielding changes and respective fielding strategies . Teams are now using Fielding data to improve their performance and strategy. 

The road ahead for Quidich Tracker

• Scaling across the entire cricket ecosystem (adding new geographies, formats, and competitions) 
•  A "Historical Data" platform provides fielding data which can be drilled down to a ball by ball level, showing the positions of 11 players at the point of release of each ball for analytics and improved team strategy and workload management.
• An integrated 'no ball' decision-making system based on field restrictions during the game of cricket. For example, during a power play, only two fielders can be outside the 30-yard circle at the ball's release point.