Player detection method based on scale attention and scale equalization algorithm.

Area of Science:

• Computer Vision
• Machine Learning
• Sports Analytics

Background:

• Object detection in team ball games faces challenges with scale variations, leading to missed detections for smaller players and reduced accuracy for larger ones.
• Existing methods often rely heavily on dataset scale statistics, which can be unreliable for diverse player sizes.

Purpose of the Study:

• To develop an improved object detection method for team ball game players that addresses scale-related inaccuracies.
• To enhance the detection of players across a wide range of bounding box sizes, particularly small and large ones.

Main Methods:

• A novel multi-scale attention mechanism is proposed, utilizing a specially created SIoU (Similar to Intersection over Union) label to represent multi-scale features.
• The SIoU label guides the training of multi-scale attention network modules at two granularity levels.
• An integrated scale equalization algorithm within SIoU labels is employed to improve detection of multi-scale targets in imbalanced datasets.

Main Results:

• Experiments on basketball, volleyball, and ice hockey datasets demonstrate significant improvements in detection accuracy.
• The proposed method improved detection accuracy for smaller bounding boxes by 11%, 15%, and 9% respectively.
• Detection accuracy for larger bounding boxes was improved by 7%, 8%, and 4% respectively.

Conclusions:

• The proposed two-fold approach effectively addresses scale-related challenges in object detection for team sports.
• The multi-scale attention mechanism and scale equalization algorithm enhance the robustness and accuracy of player detection across various scales.