The FORCE dataset provides a comprehensive set of human-object interaction scenarios, emphasizing intuitive physics. Unlike existing datasets, FORCE features detailed physical interactions with objects of varying resistances, such as pushing, pulling, and carrying. This dataset aims to bridge the gap in understanding the physical dynamics of human-object interactions.
assets/Contains scanned 3D models of all objects used in the dataset.force_npz/Includes human motion in SMPL format, along with the corresponding object motion data (pose and translation).annotations_final/Provides the annotations for all interactions (weights, objects, sequences, etc.)lib/Contains utility libraries for processing the dataset.visualize_force_dataset.pyA Python script to visualize and explore the dataset.betas.pklSMPL body shape parameters.
- Physical interactions between humans and objects with various resistance levels.
- Actions include pushing, pulling, and carrying objects.
- Install the required dependencies using a package manager of your choice (e.g., pip or conda).
- Run the visualization script:
python visualize_force_dataset.py- Load data from the
force_npz/directory to access SMPL and object motion data. - Use the
visualize_force_dataset.pyscript to visualize interactions.
If you have any questions or need assistance with the dataset, feel free to contact us: xzhang@mpi-inf.mpg.de
If you use the FORCE dataset in your research, please cite:
@inproceedings{zhang2024force,
title = {FORCE: Dataset and Method for Intuitive Physics Guided Human-object Interaction},
author = {Zhang, Xiaohan and Bhatnagar, Bharat Lal and Starke, Sebastian and Petrov, Ilya A. and
Guzov, Vladimir and Dhamo, Helisa and Pérez Pellitero, Eduardo and Pons-Moll, Gerard},
booktitle = {International Conference on 3D Vision (3DV)},
month = {March},
year = {2025},
}Thank you for your interest in the FORCE dataset!