Our paper "Bridging the gap: Doubles badminton analysis with singles-trained models" has been accepted to Signal Processing: Image Communication!
posted on July 28, 2026

Our paper “Bridging the gap: Doubles badminton analysis with singles-trained models” has been published in Signal Processing: Image Communication!

Summary

This research addresses the challenges of multi-person tracking and tactical performance analysis in doubles badminton. By proposing a novel cross-format transfer learning framework using Spatio-Temporal Graph Convolutional Networks (ST-GCN) and pose-based keypoint representations (ViT-Pose), the study enables models trained exclusively on singles match data to be effectively transferred to doubles matches without requiring massive doubles-specific retraining datasets.