This repository contains the source code and a basic demonstration notebook for our paper: "Semantially-Driven Object Search Using Partially Observed 3D Scene Graphs".
The repo has been tested to work with Python >= 3.11, so those versions are recommended. Whether any version below that will work has not been tested.
In a desired location on your computer, do:
git clone git@github.com:UW-CTRL/prophe-c.git
Change directories into the repo folder:
cd prophe-c
Then, set up a Python virtual environment:
python3 -m venv env
Activate the environment with:
source env/bin/activate
From there, install the required packages:
pip install -r requirements.txt
Lastly, you will need to download the medium and large English language datasets from SpaCy. In the same terminal, do:
python3 -m spacy download en_core_web_md
python3 -m spacy download en_core_web_lg
The file demo.ipynb is a Jupyter notebook that can be used to run a couple object-search scenarios that match the ones shown in our paper's "Qualitative Results" section. Follow the cells there, and it should all work!
If you wish to use or reference this code for your own research, please use the following bibtex! :)
@inproceedings{
remy2023semanticallydriven,
title={Semantically-Driven Object Search Using Partially Observed 3D Scene Graphs},
author={Isaac Remy and Abhishek Gupta and Karen Leung},
booktitle={NeurIPS 2023 Foundation Models for Decision Making Workshop},
year={2023},
url={https://openreview.net/forum?id=YVZ03XsMfg}
}