xasd2003/motion_map_3d

Action Recognition with Motion Map 3D Network

Action Recognition with Motion Map 3D Network

By YuChao Sun, Wennan Yu, Feiwu Yu.

The paper is publishing on the journal Neurocomputing, and can be found here

Introduction

The Motion Map 3D Network(MM3D) is consist of two network. The generation network can represent the content of a video with arbitrary video length by a motion map. The discrimination network can learn the discriminative information based on the motion maps.

Fig.1 MM3D framework

Requirements:

1. Python.
2. If you only use the generation network, you can only install C3D_v1.0.
3. If you want use both two networks, make sure that you have install both C3D_v1.0 and C3D_v1.1.

Usage:

Will be availble soon.