PROJECT TITLE :
Tangent Bundles on Special Manifolds for Action Recognition
Increasingly, machines are interacting with folks through human action recognition from video streams. Video data can naturally be represented as a third-order knowledge tensor. Although several tensor-based mostly approaches have been proposed for action recognition, the geometry of the tensor space is seldom thought to be an important side. In this paper, we tend to stress that a information tensor is connected to a tangent bundle on a special manifold. Using a manifold charting, we have a tendency to can extract discriminating information between actions. Information tensors are initial factorized using high-order singular price decomposition, where each issue is projected onto a tangent house and the intrinsic distance is computed from a tangent bundle for action classification. We examine a customary manifold charting and a few alternative chartings on special manifolds, particularly, the special orthogonal cluster, Stiefel manifolds, and Grassmann manifolds. Because the proposed paradigm frames the classification theme as a nearest neighbor based on the intrinsic distance, previous coaching makes no sense. We evaluate our method on 3 public action databases as well as the Cambridge gesture, the UMD Keck body gesture, and also the UCF sport datasets. The empirical results reveal that our methodology is highly competitive with the current state-of-the-art methods, sturdy to tiny alignment errors, and nevertheless easier.
Did you like this research project?
To get this research project Guidelines, Training and Code... Click Here