Video Interaction Recognition and Detection

This is a master thesis project aiming for interaction classification and detection in videos. So, there are two main parts in the projects, including interaction classification and interaction detection.

Software dependencies:

• Python 2 or Python 3  
• tensorFlow r1.2  
• opencv 3.2.0  
• numpy  
• xlrd  
• matplotlib.pyplot  
• imutils  

The file strcture:

• Project/
  • datasets/:   The UT-Interaction dataset (videos)  
  • log/:       The logging files which have all training and evaluating records  
  • variables/: The viarible parameter files  
  • src/
    • common/: Common files  
    • datasets/:   The python files for data pre-processing  
    • dataVisualization/: 3D CNN data visualization, and evaluating results visualization
    • model/:     3D ConvNet basic model components.  
    • train/:     Interaction classification networks, including singleNet, fullNet, pre-trained Net, etc.  
    • humanSegmentation/: person detection, tracking and segmentation
    • detector/:   interaction detection.