Break4Models Dataset

Break4Models is a small-scale 4-shape dataset for deepfracture paper. It contains simulation data of various 3D objects undergoing impact and breaking fracture pattern, designed for training and evaluating neural networks that can predict object fracture patterns for specific one target shape. It was created by FractureRB.

📖 For more details, please visit:

• GitHub Repository
• Project Page

Overview

This dataset contains simulation data of 4-shape 3D objects undergoing impact and breaking fracture pattern. Each target shape includes multiple simulation runs with different impact conditions, providing a small-scale dataset for learning relationship between impact collision condition and physics-based fracture patterns.

Dataset Structure

break4models/
├── _out_base/          # Base object simulations
├── _out_pot/           # Pot object simulations  
├── _out_squirrel/      # Squirrel object simulations
├── _out_bunny/         # Bunny object simulations
└── README.md           # This file

Each object category directory contains:

• meta.json - Metadata with simulation statistics
• obj/ - 3D mesh files (.obj format) for each simulation step
• info/ - Collision information files
• impact/ - Detailed impact data in JSON format
• initial_cond/ - Initial conditions for simulations
• nii/ - NIfTI format data (if applicable)
• gif/ - Animation files showing breaking process

Data Format

Meta Information (meta.json)

{
  "target_shape": "base",
  "max_gssdf_val": 0.128913,
  "avg_max_gssdf_val": 0.127481,
  "min_gssdf_val": -1.045852,
  "avg_min_gssdf_val": -1.045852,
  "start_time": "2024-03-30T08:58:40",
  "stop_time": "2024-03-30T10:40:07",
  "total_valid": 277
}

Impact Data (impact/*.txt)

JSON format containing collision information:

• collElems: Collision element IDs
• collPoints: Collision point coordinates [x, y, z]
• collDirections: Collision direction vectors
• collVels: Collision velocities
• collImpulse: Collision impulse values
• collForce: Collision force magnitudes

Info Files (info/*.txt)

Simple format: directory/,frame_number

Usage

This dataset is designed for:

• Training neural networks for physics-based fracture pattern
• Evaluating fracture prediction models
• Research in computational physics and computer graphics
• Development of real-time simulation systems

Citation

If you use this dataset in your research, please cite:

      @article{huang2025deepfracture,
        author = {Huang, Yuhang and Kanai, Takashi},
        title = {DeepFracture: A Generative Approach for Predicting Brittle Fractures with Neural Discrete Representation Learning},
        journal = {Computer Graphics Forum},
        pages = {e70002},
        year = {2025},
        keywords = {animation, brittle fracture, neural networks, physically based animation},
        doi = {https://doi.org/10.1111/cgf.70002},
        url = {https://onlinelibrary.wiley.com/doi/abs/10.1111/cgf.70002},
        eprint = {https://onlinelibrary.wiley.com/doi/pdf/10.1111/cgf.70002}
      }

License

This dataset is licensed under the Public Domain Dedication and License (PDDL).

Contact

For questions or issues, please open an issue on the Hugging Face dataset page.