The Center for Data Innovation recently spoke with Ismail Baran, CEO of HITAI, a London-based company that uses computer vision and machine learning to turn sports footage into detailed performance data. Baran explained how HITAI analyzes athletic movements, generates performance metrics, and helps make sports analytics more accessible to coaches, athletes, and sports organizations.
David Kertai: What does HITAI offer?
Ismail Baran: Combat sports generate enormous amounts of valuable performance data, yet coaches still often rely on observation, experience, and intuition to evaluate athletes and make decisions. Unlike sports such as football or basketball, combat sports often lack detailed, real-time data that coaches and athletes can use to assess performance objectively. Collecting this information has traditionally required wearable sensors, specialized equipment, or expensive tracking systems, limiting access for many athletes and organizations.
HITAI aims to make performance analysis more accessible by using computer vision and machine learning to analyze video footage, track athletes, recognize their movements, and convert individual actions into measurable performance metrics.
Kertai: How does HITAI track fighters and recognize their movements using footage from a single camera?
Baran: We combine several computer-vision and machine-learning techniques to analyze video. First, our system detects and tracks each athlete using footage from a single camera, such as a recording of a fight or training session. It then analyzes their movements, positioning, speed, interactions, and sequences of actions over time.
This approach is important because many athletic movements unfold across multiple video frames rather than appearing in a single image. A punch, kick, or defensive maneuver involves a sequence of actions that the system must interpret as a whole. By analyzing these sequences, we can identify specific actions, attribute them to the correct athlete, and transform video footage into structured performance data that can be measured and compared.
Kertai: What performance metrics can HITAI generate?
Baran: Depending on the sport and application, we can generate metrics such as strike counts, successful versus attempted actions, movement patterns, activity levels, action frequency, speed, estimated impact, defensive actions, combinations, target zones, and performance comparisons between rounds.
We also link each metric directly to the corresponding video footage, allowing a coach to select a statistic and immediately review the moment it represents. The value is not simply having more statistics, but rather that coaches can identify patterns that are difficult to track consistently in real time. For example, they can compare training sessions to see whether an athlete’s output declines late in a round, determine which techniques succeed most often, or measure progress throughout a training camp.
Kertai: How do you train and validate HITAI’s AI models to recognize athletes’ movements accurately?
Baran: Reliable sports analysis depends on diverse, high-quality training data. Combat sports present a particular challenge because movements vary significantly based on the athlete, fighting style, camera angle, lighting conditions, distance, and speed.
We train our models using carefully annotated video from training sessions, sparring, and competitions. Human reviewers and automated annotation tools help label actions and maintain data quality. We also use synthetic data, which consists of computer-generated examples that simulate movements and scenarios that may be difficult, rare, or expensive to capture in real-world footage.
To validate the models, we test them on video they have never encountered before. This allows us to measure how accurately they recognize movements in unfamiliar situations. As we work with more gyms, coaches, promotions, and athletes, we gain access to a broader range of real-world footage, helping us continuously improve the accuracy and robustness of our models.
Kertai: How does HITAI’s approach differ from other systems in the sports space?
Baran: Our biggest advantage is accessibility. Many sports-tracking systems require athletes to wear sensors, use specialized equipment, or compete in venues equipped with dedicated camera networks. While these systems can provide valuable information, they can also be expensive and difficult to deploy at scale.
Our approach works with existing video workflows. If a gym, federation, broadcaster, or promotion already has a suitable camera feed, our system can extract performance intelligence directly from that footage. There is no need to change what athletes wear or modify the competition environment.
For broadcasters and promoters, this also makes adoption much easier. The same video feed used for a live broadcast can become the source of real-time statistics, graphics, and performance insights. Ultimately, we want advanced sports analytics to be available not only to organizations with large technology budgets, but also to smaller promotions, coaches, gyms, and individual athletes.
