Gait Recognition
GaitGL with local temporal aggregation and global–local feature extraction
Work building on the GaitGL model, which identifies people from the way they walk rather than from facial features. The architecture pairs a global branch that captures whole-silhouette motion with local branches over horizontal body regions, combined through local temporal aggregation across the gait sequence — the global–local split is what lets it pick up both overall gait rhythm and limb-level detail.