September 23, 2025Open Access

JMM-TGT: Self-supervised 3D action recognition through joint motion masking and topology-guided transformer v1

View Full Paper

Ask AI

Bookmark

JMM-TGT model enhances action recognition accuracy by 1.5% to 7.9% across evaluation settings.
Joint Motion Masking improves subtle joint movement perception, boosting the model's performance significantly.
Integration of topological relationships within the transformer enables better capture of complex motion dynamics.
Comparison with mainstream models highlights JMM-TGT's superior action recognition capabilities.

Discussion

No takes yet. Share an insight, caveat, or question.

Self-supervised learning improves 3D action recognition in diverse datasets, indicating advances in topology-guided models.

WenHan not provided (2025) studied this question.

View Full Paper

Ask AI

Bookmark