SMART: SMPLest-X Mesh Adaptation and RAFT Tracking for Soccer Pose Estimation
Researchers have developed a new method called SMART for estimating the 3D poses of soccer players from broadcast video. This approach fine-tunes the SMPLest-X model and incorporates RAFT dense optical flow tracking, along with other enhancements like foot-plane anchoring and temporal smoothing. SMART significantly improved upon the FIFA baseline score, achieving a 38.6% enhancement on the validation set and a score of 0.593 on the test set. AI
IMPACT This method sets a new benchmark for 3D pose estimation in sports, potentially improving sports analytics and broadcast technologies.