Gesture recognition on human pose features of single images

homer_gestrec classifies five basic gesture types using OpenPose.

Gesture Id	Gesture class	Description
1	waving_right	Waving with right hand
2	waving_left	Waving with left hand
3	pointing_right	Pointing with right hand
4	pointing_left	Pointing with left hand
5	stop	Prohibition sign

Requirements

• pip install scikit-learn==0.20.0
• pip install imblearn

Usage

Load pre-trained models

Run the script for loading the pre-trained models:

sh load_models.sh

Predict gesture using trained-models

• Note the sample_features.npz must containt the extracted pose features as of now

python gesture_classification.py --pred=true --pretrained=true --model=random_forest.pkl

Train your own models

For training your own model, run

python gesture_classification.py --train=true You can change path to training data in ./config/training_data_description.yaml by modifying path_to_trainig_data paramter. Note, we expect numerical labels for classes in training dataset.

Cite

If you use this work please cite us as follows:

@inproceedings{memmesheimer2018gesture,
  title={Gesture Recognition On Human Pose Features Of Single Images},
  author={Memmesheimer, Raphael and Mykhalchyshyna, Ivanna and Paulus, Dietrich},
  booktitle={Intelligent Systems (IS), 2018 9th International Conference on},
  pages={1--7},
  year={2018},
  organization={IEEE}
}