nikoloside/deepfracture-runtime

1---2title: DeepFracture Runtime3emoji: 💥4colorFrom: gray5colorTo: red6sdk: static7pinned: false8license: mit9models:10  - nikoloside/deepfracture11tags:12  - fracture13  - physics14  - simulation15  - graphics16  - three-js17---18 19# DeepFracture — Live Web Runtime20 21Real-time, in-browser neural brittle-fracture demo: click the object to shoot22a projectile; the impact becomes the paper's 7-D collision embedding, a JS23mirror of the Siren encoder selects the nearest VQ-VAE codebook entry24([nikoloside/deepfracture](https://huggingface.co/nikoloside/deepfracture)),25and the matching pre-decoded fragment set (decoder + watershed segmentation,26baked offline per codebook entry) swaps into the rigid-body simulation.27 28Three.js + Rapier. Also live at <https://nikoloside.graphics/deepfracture-live/>.29 30## Paper31 32This demo accompanies:33 34> **DeepFracture: A Generative Approach for Predicting Brittle Fractures with35> Neural Discrete Representation Learning.** Yuhang Huang, Takashi Kanai.36> *Computer Graphics Forum*, e70002, 2025.37> [DOI: 10.1111/cgf.70002](https://doi.org/10.1111/cgf.70002)38 39```bibtex40@article{huang2025deepfracture,41  author  = {Huang, Yuhang and Kanai, Takashi},42  title   = {DeepFracture: A Generative Approach for Predicting Brittle43             Fractures with Neural Discrete Representation Learning},44  journal = {Computer Graphics Forum},45  pages   = {e70002},46  year    = {2025},47  doi     = {https://doi.org/10.1111/cgf.70002}48}49```50 51Papers: [DeepFracture (CGF 2025)](https://nikoloside.graphics/deepfracture/) ·52[Far-From-Boundary Fields (SMI 2026)](https://nikoloside.graphics/far-from-boundary-fields/) ·53Code: [TEBP-DeepFracture](https://github.com/nikoloside/TEBP-DeepFracture)54