Effective fusion method on silhouette and pose for Gait Recognition

Silhouette and pose are two common features to extract the descriptive unique patterns of a person’s gait, good performance has been already achieved driven by deep learning techniques. However, some issues still exist, silhouette is known be sensitive changes appearance while not so discriminative as even though it considered being more robust. Therefore, advantageous fuse into one model achieve both accuracy well robustnesss. In this paper, we propose simple yet effective fusion combine features, where first aligned in distribution then combined Compact Bilinear Pooling higher order fine-grained information. The superiority proposed method verified through experiments on benchmark datasets CASIA-B, OUMVLP, SOTON-small. SOTA results with an average 96.9% rank-1 accuracy. addition, cross data conducted demonstrate robustness our method.