Steps to execute for now.  


/EDA --> helper.py first for creating bags and subbags


/I3D_extraction --> extract_features.py Used  for kinetic feature extraction from the videos (bag_index.py) to use them for label generator 

/labelGenerator --> train_label_generator.py Used for training the label generator by parsing the I3D features and getting the probablity scores 

/labelGenerator --:> generate_pseudo_labels.py For creating the labels from the trained label_generator.pth on the I3d features

/Behaviour --> build_behaviour_cache Where entire video were parsed in face_landmarker.task model by google and got the behaviour cache

/graph_model --> build_graph_dataset.py to build the graph first with the features being, Behavioural , I3d and psudo labels by label generator
/graph_model --> train_gnn.py Finally train the gnn to save it as gnn_model.pth