'Bag of Poses' is a group body language analysis tool. Combining the work of OpenPose - a real-time multi-person system to jointly detect human body, hand, and facial keypoints with a bag of words approach, 'Bag of Poses' is able to analyse large group images for supervised machine learning tasks.

This is a Python implementation with TensorFlow (CUDA optimised) and scikit-learn.

To download pre-trained Coco 4000 weights: cd model; sh get_model.sh

Project Architecture :

:::python 
- root dir/
	|- images/
			|- test /
				|- obj1/
				|- obj2/

			|- train /
				|- obj1/
				|- obj2/

	|- helpers.py
	|- Bag.py 


:~$ python Bag.py --train_path images/train/ --test_path images/test/

Saves checkpoint after each labelled object for loading large data: ./checkpoint.pkl

Remove completed labelled object by running again:

:~$ python Bag.py --train_path images/train/ --test_path images/test/ --checkpoint_path ./checkpoint.pkl

Produces Confusion Matrix and Historgram of Words: