Human Activity Recognition from videos using Deep Learning on Graphically Structured Data
Code TAG - Contains the necessary MATLAB files to create a TAG from tracking data of videos
TAG Results - Contains the results of TAG Code on four datasets - ME, SBUKI, URFD, UTI
TAG2Vec.ipynb - Converts the TAG model into vectorized format to use as input to DL model
DLModel.ipynb - LSTM based architecture for Human Activity Recognition
New Formalism - Contains the python files necessary to build new formalism based on egocentric reference frame
Modified TAG - Contains results of building the New Formalism on the ME dataset