Social-Interaction-Detection-Code

Code Related to the Roman paper: "Automatic detection of human interactions from RGB-D data for social activity classification"

If you use the dataset or the code for your research, please cite our RO-MAN 2017 that describes the data collection in detail

@article{coppola2017automatic,
  title={Automatic detection of human interactions from RGB-D data for social activity classification},
  author={Coppola, Claudio and Cosar, Serhan and Faria, Diego R and Bellotto, Nicola and others},
  year={2017},
  publisher={IEEE}
}

The code relies on the usage of the UoL 3D Social Interaction Dataset described in https://lcas.lincoln.ac.uk/wp/research/data-sets-software/uol-3d-social-interaction-dataset/

The code is written in Matlab. To be able to run it, the dataset must be downloaded and the code configured.

Configuration

The config file in the K2_Code folder has to be edited with the path of four elements:

1. The folder where this repository is placed
2. Path to the data folder ( in the dataset: [...]/uol_social_interaction_dataset/social_interaction_segmentation/extrated_data)
3. Path to the csv annotation file
4. Path to where the output files are going to be placed

Run

To run the experiments, after the configuration, run the following code files:

k2_init 
k2_hmm_full_crossvalidation

Take note that the parameters of the experiments can be configured in k2_hmm_full_crossvalidation.m