TLDR: FST.ai 2.0 is an explainable AI ecosystem designed to improve decision-making in Olympic and Paralympic Taekwondo. It integrates pose-based action recognition, uncertainty modeling, and visual explanations to support referees, coaches, and athletes in real-time. The system significantly reduces decision review times (by 94.8%), increases referee trust (93%), and aids in fair Para-Taekwondo classification. It also provides interactive dashboards for performance analysis and referee training, emphasizing transparency, fairness, and human-in-the-loop control.
In the dynamic and high-stakes world of Olympic and Paralympic Taekwondo, fair, fast, and inclusive decision-making is paramount. A new explainable AI ecosystem, FST.ai 2.0, is stepping onto the mat to support referees, coaches, and athletes, aiming to bring unprecedented transparency and efficiency to the sport.
Developed by a team including Keivan Shariatmadar, Ahmad Osman, Ramin Ray, Usman Dildar, and Kisam Kim, FST.ai 2.0 is an evolution of an earlier framework. It addresses critical challenges in combat sports, such as the need for real-time action recognition, robust decision-making under uncertainty, and inclusive support for Para-athletes, aspects often lacking in existing AI systems.
What is FST.ai 2.0?
At its core, FST.ai 2.0 is a comprehensive AI system designed to act as a collaborative co-pilot, enhancing human expertise rather than replacing it. It integrates several advanced technologies:
- Pose-Based Action Recognition: Using sophisticated computer vision techniques, the system analyzes the movements of athletes in real-time, identifying specific Taekwondo actions like head kicks, punches, and defensive maneuvers.
- Uncertainty Quantification: Recognizing that not all decisions are clear-cut, FST.ai 2.0 models different types of uncertainty. This allows the system to flag borderline cases for human review, ensuring that critical decisions are made with the highest possible confidence.
- Explainability Overlays: To build trust and provide transparency, the system generates visual explanations. These overlays highlight the specific body parts or moments in a video that led the AI to its decision, making its reasoning understandable to referees, coaches, and athletes.
- Interactive Dashboards: These user-friendly interfaces provide real-time and post-event analytics for referees, coaches, and federation officials. They track performance trends, scoring latency, decision consistency, and offer valuable insights for training and strategic planning.
Key Applications and Benefits
FST.ai 2.0 offers a wide range of applications across the Taekwondo ecosystem:
- AI-Assisted Refereeing: The system provides real-time visual support to referees, significantly reducing the time spent on video reviews. During a pilot deployment, it demonstrated a 94.8% reduction in decision review time, dropping from nearly 90 seconds to just 4.6 seconds per call. Referees reported a 93% trust level in AI recommendations, and manual jury overrides decreased by 42%.
- Athlete and Coach Support: Coaches and athletes can access detailed performance analytics, including scoring probability traces, technique breakdowns, and movement consistency feedback. This data-driven approach enables personalized training plans and tactical adjustments.
- Para-Taekwondo Classification: A dedicated module assists in the fair and reproducible classification of Para-athletes. By analyzing motion-capture data and biomechanical features, it helps ensure equitable assessments, flagging ambiguous cases for expert review to minimize misclassification risks.
- Referee Training and Education: The system includes an adaptive module for referee education, offering scenario-based video replays with AI insights, personalized feedback, and skill tracking. This helps referees understand decision criteria, identify biases, and improve consistency.
Also Read:
- Enhancing Sports Science Data with Knowledge Graphs and Ontologies
- HumanCM: Accelerating Human Motion Prediction with Single-Step Generation
Ethical Considerations and Future Outlook
The developers emphasize that FST.ai 2.0 is built with strong ethical principles, focusing on fairness, transparency, and privacy. All AI-assisted decisions are logged for auditability, and human officials retain the final override authority. Data is processed under strict privacy frameworks, and the system undergoes regular bias detection and fairness audits.
The successful pilot at the 2025 World Cadet Championships in Fujairah validated the system’s real-world feasibility and value. The long-term vision for FST.ai 2.0 is to establish it as a globally adopted infrastructure, supporting national federations, coaching academies, and referee certification programs, fostering a transparent and explainable digital twin ecosystem for Taekwondo. For more detailed information, you can refer to the full research paper: FST.ai 2.0: An Explainable AI Ecosystem for Fair, Fast, and Inclusive Decision-Making in Olympic and Paralympic Taekwondo.
FST.ai 2.0 represents a significant step towards integrating trustworthy, data-driven, and human-aligned AI into sports, ensuring that the spirit of fair play is upheld with cutting-edge technology.
