Kinect Real-time (Online) Action Detection

We implemented an online action detection software that is able to take a stream of live frames from Microsoft Kinect V1 and detect the actions being performed in an online manner. The training on the dataset is done in a Matlab and the training results are suplied to this project to make the online detection. We used "MSR Action3D Dataset" in the training phase. So, the program can detect only actions provided in this dataset.

• MSR Action3D Dataset: http://research.microsoft.com/en-us/um/people/zliu/actionrecorsrc/

This project is implemented by (in Alphabetical order)

1- Abdelrahman Ali Morgan 2- Abd-Elrahman Amir Elessawy 3- Amr Saleh Mohamed Aly 4- Bassem Mohamed Abdelmoneim 5- Mohamed Hassan Khames

This is our graduation project and our supervisor is Dr.Mohamed Bassiouny

This implementation is based on the paper

[Linear-time Online Action Detection From 3D Skeletal Data Using Bags of Gesturelets] by Moustafa Meshry, Mohamed E. Hussein, Marwan Torki Dept. of Computer Science and Engineering, Egypt-Japan University of Science and Technology