TLDR: FST.ai is a new AI-powered framework designed to improve officiating in Sport Taekwondo by enhancing fairness, speed, and trust. It uses computer vision and deep learning to detect and score head kicks in real-time, reducing decision times from minutes to seconds. The system provides recommendations to human juries, maintaining human oversight, and addresses ethical concerns like data privacy and algorithmic transparency. Its core technology is adaptable to other combat and team sports, offering broader applications in performance monitoring and injury prevention.
In the fast-paced world of competitive sports, particularly in disciplines like Sport Taekwondo, fair and accurate officiating is paramount. However, traditional manual systems, even with the aid of Instant Video Replay (IVR), often struggle with delays, subjective interpretations, and inconsistent decisions. This can undermine the integrity of the competition and erode trust among athletes and fans.
A recent research paper introduces FST.ai, a groundbreaking AI-powered framework designed to significantly enhance officiating in Sport Taekwondo. The system specifically targets the challenging task of real-time head kick detection and scoring, aiming to improve fairness, speed, and trust in competitive environments. The paper, titled “AI-Powered Precision in Sport Taekwondo: Enhancing Fairness, Speed, and Trust in Competition (FST.ai)”, was authored by Dr. Keivan Shariatmadar and Dr. Ahmad Osman.
Addressing Key Challenges in Officiating
Current officiating methods, even with video replay, can lead to substantial delays. For instance, a single IVR request to review a head kick decision at a recent championship took nearly 90 seconds to resolve, halting the match and potentially influencing its outcome. Such inefficiencies can accumulate, leading to hours of operational loss daily in large tournaments. FST.ai seeks to eliminate these delays, reducing decision times from minutes to mere seconds.
How FST.ai Works
FST.ai leverages advanced technologies like computer vision, deep learning, and edge inference. The system works by first using pose estimation to detect the positions and orientations of athletes’ joints in real-time video frames. This transforms raw visual data into structured information. Next, deep learning models analyze sequences of this pose data to classify actions, distinguishing between different types of kicks, such as standard head kicks versus turning head kicks, which carry different point values.
Following action classification, the system performs impact detection, verifying if a valid physical contact occurred between the attacking athlete’s foot and the opponent’s head. It measures the deceleration of the attacking limb and the overlap between the foot and target zone. Based on the kick type and confirmed impact, FST.ai automatically suggests a score (3 or 5 points).
Crucially, FST.ai operates using edge computing, meaning the processing happens locally on high-performance hardware, eliminating the latency associated with cloud data transmission. This allows the system to deliver scoring suggestions to the review jury within 3 to 5 seconds of the action.
Human-in-the-Loop Design
While FST.ai provides rapid, high-confidence recommendations, it is designed as a human-in-the-loop system. The final decision authority remains with the human review jury. The system presents its suggestions with confidence scores and visual overlays, allowing jurors to make informed judgments. Any overrides by the human jury are recorded, providing valuable feedback for continuous improvement and retraining of the AI model, ensuring ethical oversight and adaptability.
Benefits for the Sport
The implementation of FST.ai offers numerous advantages:
- Dramatic Reduction in Review Time: Speeds up match flow significantly.
- Enhanced Accuracy and Consistency: AI is immune to fatigue or bias, ensuring uniform application of rules.
- Increased Trust Among Stakeholders: Transparent, evidence-backed decisions foster greater acceptance.
- Greater Transparency and Accountability: Every decision is logged, providing an audit trail for analysis and dispute resolution.
- Operational Scalability and Cost Efficiency: Reduces reliance on large review teams and expensive cloud infrastructure.
- Enhanced Spectator Experience: Faster decisions and potential for real-time overlays make the sport more engaging.
Ethical and Privacy Considerations
The developers of FST.ai have prioritized ethical considerations and privacy. The system processes video in real-time on local devices, meaning raw visual data is not stored or shared. Only decision-relevant metadata, such as keypoint trajectories, is retained. It also generates anonymized overlays to assist jurors, reducing exposure to raw video content. The system’s explainability features provide a rationale for each decision, preventing a ‘black box’ effect and building trust. Furthermore, models are trained on diverse datasets to ensure fairness across different demographics and prevent bias.
Also Read:
- Enhancing Medical Rehabilitation with AI: A New Framework for Motion Data Optimization and Anomaly Detection
- AI System Enhances Shoulder Fracture Detection in X-rays
Beyond Taekwondo: Extended Applications
The underlying architecture of FST.ai is highly adaptable and can be applied to a wide range of other sports. It holds promise for other combat sports like Karate, Judo, Wrestling, and Boxing, where it could detect and score strikes, throws, or grapples. It can also be customized for Para Taekwondo, accounting for diverse physical capabilities.
Beyond combat sports, the framework could assist referees in team sports like football, basketball, and handball by recognizing fouls, off-ball movements, or illegal screens. Its capabilities also extend to athlete performance monitoring, injury risk prediction, and match simulation for tactical analysis. This versatility positions FST.ai as a foundational technology for modernizing sports officiating and athlete development.
FST.ai represents a significant step forward in integrating artificial intelligence into sports officiating. By augmenting human judgment with objective, data-driven insights, it aims to uphold the principles of fairness, efficiency, and trust in competitive sports. For more details, you can refer to the research paper.
