UZY Positioning for Lithic Artifacts

Python implementation of the Grosman 2008 UZY positioning methodology for objective, reproducible orientation of 3D scanned lithic artifacts.

---

Overview

This repository provides a complete Python implementation of the UZY (Uzy Smilansky) positioning method for objectively orienting 3D lithic artifacts. The methodology uses:

• Eigenvector analysis of surface normal distribution (inertia tensor)
• Planform optimization (maximum projected area for face-on view)
• Mirror symmetry alignment (bilateral symmetry detection)
• Upright orientation (longest dimension vertical)

This eliminates researcher subjectivity in artifact measurement. The original paper demonstrated 18-28% measurement variation depending on manual positioning choices.

---

Quick Start

Installation

Option 1: UV (Recommended)

UV is a fast Python package manager.

# Setup environment
./setup-env.sh          # macOS/Linux
# or
setup-env.bat           # Windows

# Activate
source .venv/bin/activate

Option 2: Standard pip/venv

python -m venv .venv
source .venv/bin/activate  # Windows: .venv\Scripts\activate.bat
pip install -r requirements.txt

Command-Line Usage

# Process single file (default: save as <input>_positioned.glb)
python positioning.py artifact.glb

# Specify output file
python positioning.py artifact.glb -o output.glb

# UZY positioning only (skip planform/symmetry)
python positioning.py artifact.glb --uzy-only

# Save transformation metadata to JSON
python positioning.py artifact.glb --save-metadata metadata.json

# Verbose output with statistics
python positioning.py artifact.glb --verbose

# Open 3D viewer after processing
python positioning.py artifact.glb --show

# Quiet mode (errors only)
python positioning.py artifact.glb --quiet

Python API Usage

import trimesh
from positioning import positioning

# Load mesh
mesh = trimesh.load('artifact.glb')

# Apply complete positioning
v_positioned, metadata = positioning(
    mesh.vertices,
    mesh.faces,
    include_planform=True,
    include_mirror_symmetry=True
)

# Save result
mesh.vertices = v_positioned
mesh.export('artifact_positioned.glb')

# Check metadata
print(f"Planform rotation: {metadata['planform_rotation_index']}")
print(f"Symmetry angle: {metadata['symmetry_angle']}°")
print(f"Upright rotation: {metadata['upright_rotation']}")

---

Repository Structure

MeshModelRotate/
├── positioning.py              # Main implementation (678 lines, production-ready)
├── test_positioning.py         # Single-file test script
├── utilities/
│   ├── test_rotations.py       # Rotation matrix verification
│   ├── glbutils.py             # Mesh utilities
│   └── mesh2glb copy.py        # Format converter
├── 3DModels/                   # Sample GLB files (Wansunt collection)
├── MATLAB/                     # Original MATLAB implementation (reference)
├── pyproject.toml              # Python project configuration
├── requirements.txt            # Pip dependencies
├── setup-env.sh / .bat         # Environment setup scripts
├── CLAUDE.md                   # Development documentation
└── README.md                   # This file

---

Methodology: Grosman 2008 UZY Positioning

The implementation follows the methodology described in:

Grosman, L., Smikt, O., & Smilansky, U. (2008). On the application of 3-D scanning technology for the documentation and typology of lithic artifacts. Journal of Archaeological Science, 35(12), 3101-3110.

The 6-Step Pipeline

1. Center to Origin

   • Translates mesh bounding box center to (0, 0, 0)

2. UZY Positioning

   • Computes eigenvectors of surface normal distribution (inertia tensor)
   • Aligns artifact with principal axes
   • Provides initial objective orientation

3. Planform Optimization

   • Tests 3 orthogonal rotations (90° increments)
   • Selects orientation with maximum XY projected area
   • Ensures "face-on" view (lying flat)

4. Re-center to Origin

   • Translates to origin after rotation

5. Mirror Symmetry Alignment

   • Rotates 360° in 1° steps about Z-axis
   • Finds angle minimizing distance between mirrored halves
   • Ensures consistent left/right orientation

6. Final Upright Orientation

   • Identifies longest dimension
   • Rotates to make it vertical (Y-axis)
   • Ensures artifact stands upright

Performance

• UZY positioning: ~instant
• Planform optimization: ~instant (3 rotations)
• Mirror symmetry: ~1-2 seconds (360° search)
• Total: ~2 seconds per artifact

---

Coordinate System

• Right-handed coordinate system (glTF/GLB standard)
• X-axis: Width (left/right)
• Y-axis: Height (vertical, up/down)
• Z-axis: Depth (front/back)

Final output: face-on, symmetry-aligned, upright (longest dimension vertical)

---

API Reference

positioning(vertices, faces, include_planform=True, include_mirror_symmetry=True)

Complete UZY positioning workflow.

Parameters:

• vertices (np.ndarray): Nx3 array of vertex coordinates
• faces (np.ndarray): Mx3 array of triangle face indices (0-indexed)
• include_planform (bool): Enable planform optimization (default: True)
• include_mirror_symmetry (bool): Enable symmetry alignment (default: True)

Returns:

• positioned_vertices (np.ndarray): Nx3 array of final positioned vertices
• metadata (dict): Transformation information
  • center_offset: Original bounding box center
  • uzy_transform: 3x3 UZY transformation matrix
  • planform_rotation_index: Selected rotation (0=original, 1=90°X, 2=90°X+90°Y)
  • planform_areas: List of 3 projected areas tested
  • symmetry_angle: Z-rotation angle in degrees (1-360)
  • upright_rotation: Final rotation applied ('Z+90 (X→Y)', 'X+90 (Z→Y)', or 'none')

Example:

# Full positioning
v_positioned, meta = positioning(mesh.vertices, mesh.faces)

# UZY only (skip planform/symmetry)
v_uzy, meta = positioning(
    mesh.vertices,
    mesh.faces,
    include_planform=False,
    include_mirror_symmetry=False
)

---

Utilities

Format Conversion

Convert mesh files to GLB format:

# Single file
python utilities/mesh2glb.py input.ply output.glb

# Batch conversion
python utilities/mesh2glb.py  # Processes ToConvert/ directory

Supported formats: PLY, WRL, STL, OBJ, OFF, GLB, GLTF

Testing & Validation

# Test rotation matrices
python utilities/test_rotations.py

# Export to MATLAB format
python utilities/save_mat_for_matlab.py artifact.glb

---

Dependencies

Required

• trimesh ≥4.0.0 - 3D mesh processing
• scipy ≥1.11.0 - ConvexHull for projected area
• numpy ≥1.24.0 - Array operations

Optional

• pymeshlab ≥2022.2 - Additional format support (WRL, etc.)

Install all:

pip install trimesh scipy numpy pymeshlab

---

Known Issues & Limitations

Performance

• Mirror symmetry uses brute-force 360° search (could be optimized with gradient descent)

Data Format

• Input meshes should be manifold, watertight
• No validation for degenerate triangles or NaN vertices
• Face indices must be 0-indexed (Python/NumPy convention)

Coordinate System

• Assumes right-handed coordinate system (glTF standard)
• No automatic detection/conversion of coordinate system conventions

---

Reference Implementation

The original MATLAB implementation is available in MATLAB/ for reference and validation purposes. The Python implementation is recommended for production use.

See MATLAB/README.md for details on the MATLAB scripts.

---

Scientific Reference

Grosman, L., Smikt, O., & Smilansky, U. (2008). On the application of 3-D scanning technology for the documentation and typology of lithic artifacts. Journal of Archaeological Science, 35(12), 3101-3110. doi:10.1016/j.jas.2008.06.011