Human Activity Recognition

We tried to implement the following pre-processing techniques to extract person from each frame of video:

• Interactive Foreground Extraction using GrabCut Algorithm
• Background Subtraction
• Mask RCNN

Structure of the project:

• The datasets are expected to be in the structure as shown.
• The models will be with their respective names.

projectbu
|
|-- dataset
|   |-- UCF101
|   |-- HMDB51
|
|-- dependencies
|   |-- Mask_RCNN
|
|-- scripts
|
|-- ALSTM
|   |-- autoencoders.py
|   |-- main_ae.py 

• To extract the frames, run the command below:

bash scripts/frames.sh     # To extract just the frames
bash scripts/framesbr.sh   # To extract frames with Background Reduction