Gaussian Splatting Morphing Tool

A CLI and visualization tool for smoothly interpolating between two or more Gaussian Splatting Models.
It builds intelligent point correspondences and generates intermediate morphs that transition seamlessly.
Includes both command-line and real-time interactive visualizer modes.

---

🚀 Features

• Interpolate between multiple Gaussian Splatting models
• Automatic one-to-one point correspondences based on spatial and color similarity
• Spherical linear interpolation (SLERP) for rotations
• Optional real-time visualizer with interactive morphing slider
• GPU-accelerated (PyTorch) processing
• Save interpolated frames as .ply files for animation sequences

---

🧰 Installation

Clone the repository and install the base dependencies:

git clone https://github.com/feel3x/Gaussian_Splat_Morpher.git
cd Gaussian_Splat_Morpher
pip install -r requirements.txt

If you want to use the visualizer, install the extra dependencies:

pip install -r requirements_visualizer.txt

---

🧑‍💻 Usage (CLI Mode)

You can interpolate between .ply Gaussian Splat models directly from the command line.

Basic Example

python GaussianInterpolator.py -d ./models/ -o ./output/

This searches the ./models/ folder for .ply files and generates intermediate models in ./output/.

---

Specify Individual Models

python GaussianInterpolator.py -m model1.ply model2.ply model3.ply -o ./output/

---

Adjust Interpolation Settings

Option	Description	Default
--models_to_create	Number of intermediate models per pair	10
--direct_interpolation_value	Create a single interpolated model between two specific models (e.g. 1.3)	None
--spatial_weight	Weight for spatial distance during correspondence	0.7
--color_weight	Weight for color difference during correspondence	0.3
--distance_threshold	Max allowed distance for point matching	None
--batch_size	Size of point batches for matching (reduce for lower VRAM)	512
--recenter_models	Recenter all models before interpolation	False
--normalize_scales	Normalize scales of models	False

---

Example: Generate 20 Intermediate Morphs

python GaussianInterpolator.py -m bunny_A.ply bunny_B.ply -o ./morph_output --models_to_create 20

---

Example: Export a Single Interpolated Model at 0.4

python GaussianInterpolator.py -m face_1.ply face_2.ply -o ./output --direct_interpolation_value 0.4

This saves one .ply at 40% interpolation between the first and second models.

---

🎨 Real-Time Visualizer (Optional)

To explore morphs interactively:

python visualizer.py -m model1.ply model2.ply

This launches a GUI with a slider that lets you morph smoothly between loaded Gaussian Splat models in real time.

(Visualizer requires extra dependencies from requirements_visualizer.txt.)

---

🧠 How It Works

1. Loads two or more Gaussian Splat models (.ply format)
2. Builds point correspondences between consecutive models
3. Interpolates:
   • Positions, features, scales, and opacities (linearly)
   • Rotations using SLERP
4. Handles unmatched points via fade-in/out blending
5. Outputs intermediate .ply models or visualizes them in real time

---

📜 License

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

Author: Felix Hirt
Copyright © 2025

---

🌟 Acknowledgements

This project builds upon Gaussian Splatting concepts.

The visualizer uses Nerfstudio's gSplat rasterizer: GitHub

---

Enjoy morphing! 🧩