Gymnastics has always been a sport of excellence where performance relies on absolute mastery of the body, space, and time. Yet, despite strict rules and experienced judges, scoring has often been a source of controversy, particularly at major international competitions. In recent years, artificial intelligence has gradually made its way onto the floor to analyze routines with unprecedented biomechanical precision. By combining computer vision, 3D reconstruction, and deep learning models, AI is transforming gymnastics into a true science of movement, capable of objectively measuring rotations, joint angles, angular velocity, and quality of execution1.
Seeing what the human eye cannot perceive
The first AI-powered scoring systems rely on computer vision technologies capable of tracking a gymnast’s body frame by frame. Research reported by ScienceDaily has shown that algorithms can automatically detect rotations, incorrect postures, and imbalances during aerial phases2. Whereas the human eye perceives an overall impression, AI quantifies every micro-variation, such as a knee angle that is too closed upon landing or a slightly suboptimal rotation during a somersault. This ability to objectively assess performance helps reduce discrepancies in interpretation among judges and provides a stable basis for comparison between athletes.
3D reconstruction and volumetric motion capture
The true technological breakthrough lies in the three-dimensional reconstruction of movement. Studies published by the IEEE and Springer demonstrate that 3D pose estimation makes it possible to model the position of each joint with millimeter-level precision during an entire routine34. By reconstructing the body in space, AI analyzes the trajectory of the center of mass, the alignment of body segments, and the continuity of the movement. This volumetric analysis is particularly valuable for the most complex apparatus, such as the floor, balance beam, or uneven bars, where overall coordination is crucial.
AI has already been tested in official competition
Far from being merely a laboratory concept, AI is already being used in international competitions. In 2023, Reuters reported that the International Gymnastics Federation had tested a scoring assistance system during the World Championships5. Developed in partnership with Fujitsu, this technology analyzes joint angles in 3D in real time to assist judges in evaluating performance6. According to the FIG, the system significantly reduces scoring discrepancies and improves consistency across panels, without taking the final decision away from human judges.
Measuring Rotations and Angular Velocity Using Deep Learning
The evaluation of acrobatic elements relies largely on rotation and angular velocity. Research published by Elsevier and in ACM Multimedia shows that neural networks are capable of automatically measuring a gymnast’s angular momentum while in flight78. These models analyze rotation speed, height reached, and body stability to determine whether a skill has been fully executed. This approach is essential for distinguishing between movements that are visually very similar but technically different—a major challenge in a sport where difficulty and execution are sometimes determined by just a few degrees.
Technology in the Service of Fair Play
One of the key goals of AI in gymnastics is to improve fairness. According to MIT Technology Review, automated scoring systems now achieve accuracy levels close to those of experienced human judges, while offering greater consistency1. By eliminating fatigue, cognitive biases, and variations in interpretation, AI contributes to a more consistent evaluation of performances. For federations, this technology represents a tool for global standardization, capable of ensuring that the same routine will be judged equivalently in Tokyo, Paris, or Los Angeles.
Ethical Issues: Preserving the Human Dimension of Judgment
However, the introduction of AI into arbitration raises sensitive issues.
• risk of reducing gymnastics to a set of biomechanical metrics,
• concern about a loss of artistic and expressive recognition,
• excessive reliance on algorithms in major sporting decisions,
• transparency of models and understanding of the criteria used.
Sports authorities emphasize a fundamental point: AI must remain a support tool, not an autonomous referee. The final decision must continue to incorporate human expertise, which is capable of contextualizing performance beyond mere statistics9.
Toward Data-Enhanced Gymnastics
By transforming every routine into a set of actionable data, AI is redefining how gymnastics is understood, judged, and taught. It provides coaches with detailed analytical tools to correct errors, judges with objective guidance to ensure consistency in scoring, and athletes with a more precise understanding of their own movements. Gymnastics remains a sport of aesthetics and mastery, but under the watchful eye of AI, it also becomes a science of movement, where algorithmic precision illuminates human excellence without replacing it.
Learn more
This AI-powered motion analysis now extends beyond vision systems and external sensors. For more on this topic, check out our article “AI Blends into Textiles: A Second Skin That Reveals Every Athletic Movement”, which explores how smart textiles incorporating AI enable continuous and precise tracking of movements, opening up new possibilities for training, performance, and injury prevention.
References
1. MIT Technology Review. (2022). AI judges gymnastics routines with near-human accuracy.
https://www.technologyreview.com
2. ScienceDaily. (2022). Computer vision system developed to assist with gymnastics scoring.
https://www.sciencedaily.com
3. IEEE TPAMI. (2023). Pose estimation and 3D reconstruction for gymnastics performance analysis.
https://ieeexplore.ieee.org
4. Springer. (2023). AI-based body joint estimation for gymnastics performance.
https://link.springer.com
5. Reuters. (2023). Artificial intelligence assists judges at the World Gymnastics Championships.
https://www.reuters.com
6. International Gymnastics Federation. (2023). FIG partners with Fujitsu to develop an AI-based scoring system.
https://www.gymnastics.sport /a>
7. Elsevier. (2023). Deep learning for automatic detection of angular momentum in acrobatic sports.
https://www.sciencedirect.com
8. ACM Multimedia. (2022). Automatic skill evaluation in parkour and acrobatics.
https://dl.acm.org
9. ESPN. (2024). Sports federations turn to AI to improve judging accuracy.
https://www.espn.com
