This repository provides the official implementation for our papers, which introduce novel metrics for quantifying occlusion in multi-person 3D scenes. We provide the code to compute our proposed metrics—Occlusion Index (OI), Silhouette Occlusion Index (SOI), and Crowd Index (CI)—on the 3DPW, AGORA, and OCHuman datasets.
If you use this code or our metrics in your research, please cite the relevant papers:
@inproceedings{girgin2024detection,
title={Detection and Quantification of Occlusion for 3D Human Pose and Shape Estimation},
author={Girgin, Emre and G{\"o}kberk, Berk and Akarun, Lale},
booktitle={International Conference on Pattern Recognition},
pages={368--382},
year={2024},
organization={Springer}
}
@inproceedings{girgin2023novel,
title={A novel occlusion index},
author={Girgin, Emre and G{\"o}kberk, Berk and Akarun, Lale},
booktitle={2023 31st Signal Processing and Communications Applications Conference (SIU)},
pages={1--4},
year={2023},
organization={IEEE}
}Pre-computed benchmark is available in the /results folder for direct comparison.
- Python 3.x
- A Python 3 compatible version of SMPL. We recommend egirgin/smpl_basics.
Install the dependencies.
pip install -r requirements.txtDownload the following datasets and dataset_config variable of the corresponding dataset processing script.
Important: Before running, ensure all file paths in the scripts are updated to point to your dataset and SMPL model locations.
The code is organized by dataset.
- Generate SMPL models:
python 3dpw/generate_smpl/generate-smpl.py - Calculate SOI:
python 3dpw/render_smpl/silhouette_occlusion_index.py - Calculate OI & CI:
python 3dpw/src/new_main.py - Run Experiments:
python 3dpw/src/experiments/experiments.py
- Preprocess:
python agora/preprocess.py - Calculate SOI:
python agora/regional_occlusion.py - Calculate OI & CI:
python agora/new_main.py - Run Experiments:
python agora/experiments/experiments.py
- Unzip Data:
unzip ochuman/gtSMPL.zip - Calculate OI & CI:
python ochuman/src/new_main.py(SOI is not applicable) - Run Experiments:
python ochuman/src/experiments/experiments.py