Search papers, labs, and topics across Lattice.
3
0
4
Separating spatial and temporal modeling in action recognition leads to significant performance gains, challenging the effectiveness of existing implicit coupling methods.
DualAlign boosts action quality assessment accuracy by over 21% through innovative multi-modal fusion and adaptive alignment techniques.
Adversarial attacks can now be executed on human action recognition systems without compromising motion quality, exposing critical vulnerabilities in their robustness.
