BaliMask3D Dataset & Framework

📌 Overview

BaliMask3D is a dataset and framework designed to advance 3D object completion and reconstruction, specifically for *traditional Balinese masks. These masks, which hold deep cultural and artistic significance, have been digitized to support research in *3D reconstruction, deep learning, and cultural preservation.

This project employs 360-degree photogrammetry, Signed Distance Fields (SDFs), and transformer-based architectures (e.g., VQ-VAE, SDFusion) to reconstruct incomplete 3D data with high accuracy.

🔥 Features

• 📊 *High-Fidelity 3D Dataset: Traditional Balinese masks scanned using *structured light scanning & photogrammetry.
• 🤖 *Transformer-Based 3D Completion: Implements *VQ-VAE and SDFusion models for reconstructing missing 3D data.
• 🎭 *Cultural Heritage Digitalization: Contributes to preserving *Balinese artistic heritage through digitization.
• 📏 *Evaluation Metrics: Assesses reconstructed masks using *Uniform Hausdorff Distance (UHD) and Topological Morphology Distance (TMD).
• 🏗 Flexible Framework: Includes data preprocessing, training pipelines, and evaluation scripts.

📂 Dataset Structure

BaliMask3D/
│── data/
│   ├── raw/                  # Raw 3D scans of Balinese masks
│   ├── processed/            # Preprocessed & normalized 3D models
│   ├── sdf/                  # Signed Distance Fields representations
│   ├── annotations/          # Metadata and labels
│
│── models/
│   ├── vqvae/                # VQ-VAE model implementation
│   ├── sdfusion/             # SDFusion architecture for 3D completion
│   ├── checkpoints/          # Pretrained model weights
│
│── scripts/
│   ├── preprocess.py         # Mesh normalization & SDF generation
│   ├── train.py              # Model training pipeline
│   ├── evaluate.py           # Performance evaluation (UHD, TMD)
│
│── notebooks/                # Jupyter notebooks for visualization & analysis
│── README.md                 # Project documentation

🚀 Installation

bash

Clone repository

git clone https://github.com/TrianCode/Bali3DMask.git cd Bali3DMask

Create virtual environment

python -m venv balimask3d_env source balimask3d_env/bin/activate # For Linux/macOS balimask3d_env\Scripts\activate # For Windows

Install dependencies

pip install -r requirements.txt

🔧 Usage

1️⃣ Preprocess Data

Convert raw 3D scans into normalized meshes and SDF representations. bash python scripts/preprocess.py --input data/raw --output data/processed

2️⃣ Train Model

Train VQ-VAE or SDFusion for 3D reconstruction. bash python scripts/train.py --model vqvae --epochs 100 --batch_size 16

3️⃣ Evaluate Performance

Assess reconstruction accuracy using UHD & TMD metrics. bash python scripts/evaluate.py --model vqvae --dataset data/processed

4️⃣ Visualize 3D Models

Use Jupyter Notebooks for visualization. bash jupyter notebook notebooks/visualize_results.ipynb

📊 Evaluation Metrics

• Uniform Hausdorff Distance (UHD): Measures the geometric similarity between ground truth and reconstructed 3D models.
• Topological Morphology Distance (TMD): Evaluates the structural consistency of completed models.

🤝 Contributing

We welcome contributions! Feel free to open an issue or submit a pull request:

1. Fork the repo and clone it locally.
2. Create a new feature branch (git checkout -b feature-branch).
3. Commit changes (git commit -m "Add new feature").
4. Push to the branch (git push origin feature-branch).
5. Open a Pull Request.

📜 License

This project is licensed under the MIT License. See the LICENSE file for details.

💡 Acknowledgments

This project is supported by Institut Teknologi Sepuluh Nopember (ITS) and contributors dedicated to cultural preservation and 3D research.

📧 Contact: 7025231008@student.its.ac.id | anny@its.ac.id | hadziq@its.ac.id