Compact and Discriminative Descriptor Inference Using Multi-Cues

摘要: Feature descriptors around local interest points are widely used in human action recognition both for images and videos. However, each kind of describes the characteristics reference point only from one cue. To enhance descriptive discriminative ability multiple cues, this paper proposes a descriptor learning framework to optimize at source by projection descriptors’ spaces new Euclidean space. In space, cues different fused complemented other. order make more discriminative, we learn multi-cue minimization ratio within-class scatter between-class scatter, therefore, projected is enhanced. experiment, evaluate our on tasks still Experimental results two benchmark image video data sets demonstrate effectiveness better performance method.