New gait recognition framework fuses body shape and locomotion dynamics

Researchers have developed a new gait recognition framework using deep residual networks and multi-branch feature fusion to improve accuracy in surveillance and security applications. The system employs HRNet for skeletal keypoint estimation and a ResNet-50 backbone to extract features from body proportion, gait velocity, and skeletal motion. A novel Multi-Branch Feature Fusion module, inspired by channel-wise attention, dynamically weights these features. Experiments on the CASIA-B benchmark showed a Rank-1 accuracy of 94.52% under normal walking conditions, outperforming existing skeleton-based methods for coat-wearing scenarios. AI

IMPACT Enhances biometric security capabilities with improved accuracy in challenging conditions.

RANK_REASON This is a research paper detailing a new method for gait recognition.

Read on arXiv cs.CV →

• arXiv
• ResNet-50

• paper
• other

AI-generated summary · Google Gemini · from 1 sources. How we write summaries →