phd_utils

A repository of helper functions for the various projects done during my PhD.

Usage

Clone and install this repository using:

git clone https://github.com/souljaboy764/phd_utils
cd phd_utils
pip install -r requirements.txt
pip install -e .

Dataset preprocessing

Currently the Buetepage dataset and the NuiSI dataset are preprocessed and stored in the data_preproc folder. To manually preprocess the data, please run:

git submodule update --init
python data_raw/nuisi_preproc.py --src data_raw/nuisi_dataset/ --dst data_preproc/nuisi # for the NuiSI dataset
python data_raw/buetepage_preproc.py --src data_raw/human_robot_interaction_data/ --dst data_preproc/buetepage # for the Buetepage HHI dataset
python data_raw/buetepage_preproc.py --src data_raw/human_robot_interaction_data/ --dst data_preproc/buetepage_hr --robot # for the Buetepage HRI dataset

For the Handovers Dataset, it is recommended to use the preprocessed data, as the original dataset is very large (8.9 GB compressed, 18.6GB uncompressed) To preprocess the data, first download the dataset and unzip the file Bimanual Handovers Dataset.zip to a suitable location . Assuming that the unzipped folder is at ~/Documents/Bimanual Handovers Dataset/, to preprocess the data, run:

python data_raw/alap_preproc.py --src ~/Documents --dst data_preproc/alap/ # dataset for the results reported in the paper
python data_raw/alap_preproc.py --src ~/Documents --dst data_preproc/alap_kobo/ --robot # dataset for training the model to be executed on the Kobo robot

Currently only the case of Robot-to-Human handover is explored