GaussianCube Construction

This repository contains the implementation of GaussianCube construction in the paper "GaussianCube: A Structured and Explicit Radiance Representation for 3D Generative Modeling". This repository is based on the 3D Gaussian Splatting, thanks to the authors for their great work.

Cloning the Repository

The repository contains submodules, thus please check it out with

# SSH
git clone git@github.com:/GaussianCube_Construction.git --recursive

or

# HTTPS
git clone https://github.com/GaussianCube/GaussianCube_Construction.git --recursive

Overview

The codebase has 4 main components:

• The proposed densification-constrained fitting
• Gaussian structuralization via optimal transport

Construction

We use PyTorch and CUDA extensions in a Python environment to produce trained models.

Hardware Requirements

• CUDA-ready GPU with Compute Capability 7.0+
• 24 GB VRAM (to train to paper evaluation quality)
• Please see FAQ for smaller VRAM configurations

Software Requirements

• Conda (recommended for easy setup)
• C++ Compiler for PyTorch extensions (we used Visual Studio 2019 for Windows)
• CUDA SDK 11 for PyTorch extensions, install after Visual Studio (we used 11.8, known issues with 11.6)
• C++ Compiler and CUDA SDK must be compatible

Setup

Local Setup

Our default, provided install method is based on Conda package and environment management:

SET DISTUTILS_USE_SDK=1 # Windows only
conda env create --file environment.yml
conda activate gaussian_splatting

Please note that this process assumes that you have CUDA SDK 11 installed, not 12. For modifications, see below.

Tip: Downloading packages and creating a new environment with Conda can require a significant amount of disk space. By default, Conda will use the main system hard drive. You can avoid this by specifying a different package download location and an environment on a different drive:

conda config --add pkgs_dirs <Drive>/<pkg_path>
conda env create --file environment.yml --prefix <Drive>/<env_path>/gaussian_splatting
conda activate <Drive>/<env_path>/gaussian_splatting

Modifications

If you can afford the disk space, we recommend using our environment files for setting up a training environment identical to ours. If you want to make modifications, please note that major version changes might affect the results of our method. However, our (limited) experiments suggest that the codebase works just fine inside a more up-to-date environment (Python 3.8, PyTorch 2.0.0, CUDA 12). Make sure to create an environment where PyTorch and its CUDA runtime version match and the installed CUDA SDK has no major version difference with PyTorch's CUDA version.

Running

Densification-constrained Fitting

To run the densification-constrained fitting, we provide an example below.

First, download the example data from huggingface. Then, unzip the file by the following command:

unzip example_data.zip

Then, run the following command to perform densification-constrained fitting:

```shell
python train.py -s ./example_data -m ./output_dc_fitting/ --dataset_type shapenet --white_background --test_iterations 7000 30000 --densification_interval 50 --sh_degree 0 --N_max_pts 32768

Command Line Arguments for train.py

--source_path / -s

Path to the source directory containing a COLMAP or Synthetic NeRF data set.

--model_path / -m

Path where the trained model should be stored (output/<random> by default).

--images / -i

Alternative subdirectory for COLMAP images (images by default).

--eval

Add this flag to use a MipNeRF360-style training/test split for evaluation.

--resolution / -r

Specifies resolution of the loaded images before training. If provided 1, 2, 4 or 8, uses original, 1/2, 1/4 or 1/8 resolution, respectively. For all other values, rescales the width to the given number while maintaining image aspect. If not set and input image width exceeds 1.6K pixels, inputs are automatically rescaled to this target.

--data_device

Specifies where to put the source image data, cuda by default, recommended to use cpu if training on large/high-resolution dataset, will reduce VRAM consumption, but slightly slow down training. Thanks to HrsPythonix.

--white_background / -w

Add this flag to use white background instead of black (default), e.g., for evaluation of NeRF Synthetic dataset.

--sh_degree

Order of spherical harmonics to be used (no larger than 3). 3 by default.

--convert_SHs_python

Flag to make pipeline compute forward and backward of SHs with PyTorch instead of ours.

--convert_cov3D_python

Flag to make pipeline compute forward and backward of the 3D covariance with PyTorch instead of ours.

--debug

Enables debug mode if you experience erros. If the rasterizer fails, a dump file is created that you may forward to us in an issue so we can take a look.

--debug_from

Debugging is slow. You may specify an iteration (starting from 0) after which the above debugging becomes active.

--iterations

Number of total iterations to train for, 30_000 by default.

--ip

IP to start GUI server on, 127.0.0.1 by default.

--port

Port to use for GUI server, 6009 by default.

--test_iterations

Space-separated iterations at which the training script computes L1 and PSNR over test set, 7000 30000 by default.

--save_iterations

Space-separated iterations at which the training script saves the Gaussian model, 7000 30000 <iterations> by default.

--checkpoint_iterations

Space-separated iterations at which to store a checkpoint for continuing later, saved in the model directory.

--start_checkpoint

Path to a saved checkpoint to continue training from.

--quiet

Flag to omit any text written to standard out pipe.

--feature_lr

Spherical harmonics features learning rate, 0.0025 by default.

--opacity_lr

Opacity learning rate, 0.05 by default.

--scaling_lr

Scaling learning rate, 0.005 by default.

--rotation_lr

Rotation learning rate, 0.001 by default.

--position_lr_max_steps

Number of steps (from 0) where position learning rate goes from initial to final. 30_000 by default.

--position_lr_init

Initial 3D position learning rate, 0.00016 by default.

--position_lr_final

Final 3D position learning rate, 0.0000016 by default.

--position_lr_delay_mult

Position learning rate multiplier (cf. Plenoxels), 0.01 by default.

--densify_from_iter

Iteration where densification starts, 500 by default.

--densify_until_iter

Iteration where densification stops, 15_000 by default.

--densify_grad_threshold

Limit that decides if points should be densified based on 2D position gradient, 0.0002 by default.

--densification_interval

How frequently to densify, 50 (every 50 iterations) by default.

--opacity_reset_interval

How frequently to reset opacity, 3_000 by default.

--lambda_dssim

Influence of SSIM on total loss from 0 to 1, 0.2 by default.

--percent_dense

Percentage of scene extent (0--1) a point must exceed to be forcibly densified, 0.01 by default.

--dataset_type

Type of dataset to use, shapenet by default. We provide loaders for shapenet (ShapeNet), omni (OmniObject3D) and objaverse (Objaverse) datasets.

--N_max_pts

Maximum number of Gaussians when fitting a single object , 32768 by default.

--no_tqdm

Flag to disable tqdm progress bar.

We also provide a script to perform large-scale fitting in parallel:

python run_fitting.py --source_path ./example_data --output_path ./output_dc_fitting/ --txt_file ./example_data/shapenet_car.txt --start_idx 0 --end_idx 1 --num_gpus 1 --N_max_pts 32768

Command Line Arguments for run_fitting.py

--source_path

Path to the source directory of data set.

--output_path

Path where the trained model should be stored (output/<random> by default).

--txt_file

Index file of the object to fit.

--start_idx

Starting index of the object to fit.

--end_idx

Ending index of the object to fit.

--num_gpus

Number of GPUs to use for parallel fitting, 1 by default.

--N_max_pts

Maximum number of Gaussians when fitting a single object , 32768 by default.

Gaussian Structuralization via Optimal Transport

After obtaining the fitted Gaussians, we further structuralize them via optimal transport. The running script is as follows:

python scripts/ot_structuralization.py --source_root ./output_dc_fitting/ --save_root ./output_gaussiancube/ --txt_file ./example_data/shapenet_car.txt --start_idx 0 --end_idx 1 --num_workers 1

Command Line Arguments for scripts/ot_structuralization.py

--source_root

Root path to fitted Gaussians.

--save_root

Root path to save the structuralized GaussianCubes.

--txt_file

Index file of the object to be structuralized.

--start_idx

Starting index of the object to be structuralized.

--end_idx

Ending index of the object to be structuralized.

--num_workers

Number of workers to use for parallel structuralization, 1 by default.

--bound

Bounding box of object, 0.45 by default.

--visuzalize_mapping

Flag to visualize the mapping between Gaussians and voxel grid.

The structuralized GaussianCubes after activation function are saved in ```save_root/volume_act```, which is the input for our 3D generative modeling. The non-activated structuralized GaussianCubes are saved in ```save_root/volume```. The optional visualizations of the mapping between Gaussians and voxel grid are saved in ```save_root/point_cloud```.

Citation

If you find our work useful in your research, please consider citing:

@article{zhang2024gaussiancube,
  title={GaussianCube: Structuring Gaussian Splatting using Optimal Transport for 3D Generative Modeling},
  author={Zhang, Bowen and Cheng, Yiji and Yang, Jiaolong and Wang, Chunyu and Zhao, Feng and Tang, Yansong and Chen, Dong and Guo, Baining},
  journal={arXiv preprint arXiv:2403.19655},
  year={2024}
}