SELFI3-D: Facial Mesh Reconstruction and Tattoo Visualization

SELFI3-D is a pipeline for facial mesh reconstruction and tattoo visualization from sparse input views. This project helps users visualize what they would look like with face tattoos before making permanent decisions.

Overview

Using a combination of gaussian splatting, poisson reconstruction, and texture mapping, SELFI3-D creates realistic 3D facial reconstructions from a small set of input images. The system then allows users to apply custom tattoo designs to their facial model with realistic texture mapping.

Features

• 3D facial reconstruction from sparse image views
• High-quality texture mapping of facial features
• Tattoo visualization on reconstructed face models
• Realistic rendering with proper lighting and shading

Installation

Prerequisites

• Python 3.8+
• Required packages (install via pip):

  pip install -r requirements.txt
  

Setup

1. Clone the repository:

   git clone https://github.com/jameson-crate/SELFI3-D.git
   cd SELFI3-D
   

2. Install dependencies:

   pip install -r requirements.txt
   

Usage

1. Prepare your input images in the input/ directory
2. Run the reconstruction pipeline:

   python reconstruct.py --input input/images/ --output output/mesh/
   

3. Apply a tattoo design:

   python apply_tattoo.py --mesh output/mesh/face.obj --tattoo designs/tattoo.png --output output/final/
   

Team Members

• Jameson Crate
• Michelle Chen
• Valerie Li
• Akshaan Ahuja

Project Website

For more information, visit our project website.

License

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