In recent years, the integration of artificial intelligence technologies has significantly advanced the effectiveness of athletic training programs in baseball. A critical focus is the analysis of the throw-out moment during a pitch, which is essential for optimizing pitching techniques. The analysis of the throw-out moment during a pitch is crucial for the refinement of pitching techniques. This research employs computer vision and deep learning methodologies to develop a system recognizing the baseball throw-out moment. This system enables pitchers to accurately identify their release points, allowing coaches to derive precise training cues. Additionally, our vision-based system is designed to capture essential regions for further analysis of arm slot, spin rate, and trajectory approximation and reconstruction, contributing significantly to advanced pitching strategy development.