PhysGaussian: Physics-Integrated 3D Gaussians for Generative Dynamics
Paper • 2311.12198 • Published • 22
camera_model stringclasses 1
value | w int64 1.92k 1.92k | h int64 1.08k 1.08k | fl_x float64 1.77k 1.8k | fl_y float64 1.77k 1.8k | cx float64 961 966 | cy float64 538 546 | k1 float64 0.19 0.29 | k2 float64 -1.74 -0.94 | k3 float64 1.54 3.25 | k4 float64 0 0 | p1 float64 -0 -0 | p2 float64 -0 0 | camera_angle_x float64 0.98 1 | frames listlengths 181 226 |
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
OPENCV | 1,920 | 1,080 | 1,791.146511 | 1,791.146511 | 965.898249 | 537.930464 | 0.210683 | -1.072637 | 1.815583 | 0 | -0.001161 | -0.000747 | 0.984015 | [
{
"file_path": "images/frame_00001.png",
"transform_matrix": [
[
-0.2787048635614106,
-0.1829335265758782,
0.9427931217403016,
0.31406775458273867
],
[
0.9603118944401777,
-0.04166741489654831,
0.27579875626767464,
0.030811804... |
OPENCV | 1,920 | 1,080 | 1,795.69192 | 1,795.69192 | 964.697672 | 545.604122 | 0.190044 | -0.955764 | 1.589178 | 0 | -0.001077 | 0.001212 | 0.981906 | [
{
"file_path": "images/frame_00001.png",
"transform_matrix": [
[
0.9864127047208446,
0.10760621951662103,
-0.12414092853568441,
-0.03167516591383704
],
[
-0.12428156153216843,
-0.005423978605443893,
-0.9922321747312339,
-0.059... |
OPENCV | 1,920 | 1,080 | 1,783.230009 | 1,783.230009 | 960.917222 | 542.751193 | 0.291872 | -1.74118 | 3.254339 | 0 | -0.001635 | -0.00014 | 0.987708 | [{"file_path":"images/frame_00001.png","transform_matrix":[[-0.2635309299561395,-0.04412294872212605(...TRUNCATED) |
OPENCV | 1,920 | 1,080 | 1,766.481482 | 1,766.481482 | 965.935646 | 540.414755 | 0.200272 | -0.944413 | 1.542302 | 0 | -0.000399 | 0.000419 | 0.995606 | [{"file_path":"images/frame_00001.png","transform_matrix":[[0.9506493918049368,0.18948123499522931,-(...TRUNCATED) |
Squishy 3DGS Assets (Test)
Note: These are test assets for development and evaluation of the squishy physics simulation pipeline.
Pre-trained 3D Gaussian Splatting (3DGS) scenes used by the squishy physics simulation project. Each .zip contains a full gaussian-splatting-format output directory (PLY point cloud at multiple iterations, cameras, images).
Asset Origins
All 3DGS scene assets in this repository are open-source assets from PhysGaussian (Xie et al., CVPR 2024), mirrored here to the Hugging Face Hub for convenient access. We did not train these — full credit goes to the PhysGaussian authors.
Citation:
@inproceedings{xie2024physgaussian, title = {PhysGaussian: Physics-Integrated 3D Gaussians for Generative Dynamics}, author = {Xie, Tianyi and Zong, Zeshun and Qiu, Yuxing and Li, Xuan and Feng, Yutao and Yang, Yin and Jiang, Chenfanfu}, booktitle = {Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR)}, year = {2024} }Project page: https://xpandora.github.io/PhysGaussian/ Paper: https://arxiv.org/abs/2311.12198 GitHub: https://github.com/XPandora/PhysGaussian
Contents
| File | Asset | Description |
|---|---|---|
bread-trained.zip |
Bread | Loaf of bread |
ficus_whitebg-trained.zip |
Ficus | Potted ficus plant |
pillow2sofa_whitebg-trained.zip |
Pillow on Sofa | Pillow resting on sofa |
plane-trained.zip |
Plane | Flat reference plane |
vasedeck_whitebg-trained.zip |
Vase on Deck | Ceramic vase on wooden deck |
wolf_whitebg-trained.zip |
Wolf Plush | Wolf stuffed animal |
Usage with squishy
from squishy.asset_registry import get_ply
ply_path = get_ply("bread") # extracts zip on first use
Assets are fetched automatically from this dataset repo via squishy.hub if not present locally. Set SQUISHY_HF_DATASET_REPO=deformsuite/squishy-assets in your .env.
License
The 3DGS scene assets (bread, ficus, pillow2sofa, plane, vasedeck, wolf) are open-source assets released alongside the PhysGaussian paper (CVPR 2024). No explicit license file is present in the PhysGaussian repository; these assets are redistributed here under CC BY 4.0 consistent with standard academic paper releases. Please cite the paper if you use these assets.
- Downloads last month
- 1,505