Pose Splatter: A 3D Gaussian Splatting Model for Quantifying Animal Pose and Appearance [NeurIPS 2025]

• Authors: Jack Goffinet*, Youngjo Min*, Carlo Tomasi, David Carlson (* denotes equal contribution)

Code for "Pose Splatter: A 3D Gaussian Splatting Model for Quantifying Animal Pose and Appearance."

Usage Steps

1. Collect up direction:

   python estimate_up_direction.py config.json

2. Get center and rotations:

   python calculate_center_rotation.py config.json

3. Get volume sum, decide volume_idx:

   python calculate_crop_indices.py config.json

4. Write images in HDF5 format:

   python write_images.py config.json

5. Copy images to ZARR for training:

   python copy_to_zarr.py path/to/input.h5 path/to/output.zarr

6. Train a model:

   python train_script.py config.json

7. Evaluate model:

   python evaluate_model.py config.json

8. Render an image:

   python render_image.py config.json

9. Calculate visual features:

   python calculate_visual_features.py config.json

10. Calculate visual embedding:

    python calculate_visual_embedding.py config.json

Project Checklist

• Code on GitHub
• Camera-ready on arXiv
• Add links to data
• Add more detailed usage