Abstract—Along with the development of volleyball match video analysis, this paper introduces a novel human target detection method based on an optimized You Only Look Once (YOLO) deep learning strategy. Firstly, shot segmentation (identifying scene boundaries) is performed on the volleyball videos to extract key frames. Then, using semantic annotation techniques (classifying segments to filter non-game content), the videos are described as sequences of shots composed of long shots, medium shots, close-ups, replays, and off-court shots. Secondly, to address the issue of slow speed in target detection algorithms, the backbone network of YOLOv8s is optimized by implementing lightweighting through the GhostNet network and enhancing semantic information with the Convolutional Block Attention Module (CBAM) module to improve model accuracy. Experimental results on the PascalVisual Object Classes (PASCAL VOC), Common Objects in Context (COCO), and the Volleyball datasets, as well as real volleyball match videos, demonstrate that the proposed algorithm achieves a 1.5% higher mAP@50 with 64.2% fewer computational load Giga Floating- point Operations Per Second (GFLOPs) compared to the baseline YOLOv8s, achieving an optimal balance between accuracy and efficiency, making it suitable for real-time tactical analysis and automated player performance statistics in coaching and broadcasting.
 
Keywords—semantic annotation, object detection, deep learning, lightweight network, volleyball video


Copyright © 2026 by the authors. This is an open access article distributed under the Creative Commons Attribution License which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited (CC BY 4.0).