Performance analysis in fencing is still largely dependent on manual observation and subjective interpretation by coaches and referees, which limits scalability, reproducibility, and data-driven decisionmaking. This work presents a computer vision–based system designed to automatically analyze fencing bouts from video data. The system applies human pose estimation to track athletes’ movements and extracts kinematic features such as displacement, velocity, vertical position of the weapon’s guard, and movement intensity. These features are processed to generate performance metrics and structured visualizations. The results indicate that automated video-based analysis can reduce subjectivity, increase analytical consistency, and serve as a practical support tool for athletes and coaches in fencing training and performance evaluation.