Competition
WASF First AI Taekwondo Poomsae Invitational
The inaugural WASF AI Taekwondo Poomsae Invitational Tournament employs an AI-assisted judging system to score traditional Poomsae routines, with judging criteria designed based on the official Poomsae competition rules of World Taekwondo. The following is a summary of the competition regulations.
Competition Categories: Individual, Pairs (one male, one female athlete), and Team — following the category structure used in official World Taekwondo Poomsae competition.
Age Divisions: Kids (ages 8-11), Cadet (ages 12-14), Junior (ages 15-17), Senior (ages 18 and above) — adapted from World Taekwondo's age-division conventions.
Performance Duration: Up to 90 seconds per individual/pair/team performance, consistent with Recognized Poomsae timing rules.
Scoring: Following World Taekwondo's official framework, each performance is scored out of 10.0 points, split between Accuracy (4.0 points — correct stances, correct hand/foot technique, balance) and Presentation (6.0 points — power and speed, rhythm and tempo, contrast, expression of energy). A panel of judges scores each performance; where a five-judge panel is used, the highest and lowest scores are dropped before averaging, per standard World Taekwondo practice.
WASF AI-Assisted Scoring Layer: In addition to the human judging panel above, each performance is captured using a WASF-certified front camera (technique and stance detail) and an overhead camera (footwork, stance-position accuracy, and return-to-starting-point verification), processed through an on-site edge-computing box. This produces a supplementary AI reference score covering stance accuracy, kick height, rhythm, and balance — intended to support, not replace, the human judging panel during this invitational phase. All video and skeletal data is processed on the edge-computing box on-site and is not retained or transmitted beyond the abstracted scoring result, consistent with WASF's Privacy Policy.
Note: The AI-assisted scoring engine's full technique library (stance-specific angle thresholds, trajectory-matching against standard pattern templates, gaze tracking, and start/end position fencing) is still under active development. This invitational uses the engine's current capabilities and is itself part of WASF's real-world accuracy testing process.