This is test code for a Computational Model for Human Action Understanding. Here we have considered two types of actions namely chasing and pushing in the thematic role understanding scenarios. We proposed some motion-based heuristics to model these two actions. For details, please see our paper.
- python 2.7 or might work for python3.x (not tested)
- numpy
- matplotlib
- opencv
- scikit-learn
This code has various dependencies, some are listed in Prerequisites section. The code first will generate some video dataset (synthetic) for chasing and pushing actions. Then test data based on our heuristic-based model.
Please see the Example section for different parameter combination of two actions.
To see different parameters option, from a shell (Bash, Bourne, or else) run:
python2.7 ss_action_data_classification.py -hTo test the model on default parameters values (say you want to test the model for 'num_video_files=30' videos for each action), from a shell (Bash, Bourne, or else) run the following script:
python2.7 ss_action_data_classification.py --num_video_files 30