---

[WACV 2025] Deep Geometric Moments Promote Shape Consistency in Text-to-3D Generation

This repository contains the official implementation of the paper:
Deep Geometric Moments Promote Shape Consistency in Text-to-3D Generation
📄 Paper | 🌐 Project Page

---

🛠️ Prerequisites

• Python 3.11.5
• CUDA 11.8

Install dependencies:

pip install -r requirements.txt

---

⚙️ Build Gaussian Splatting Extension

cd gs
./build.sh

---

📥 Model Checkpoint

Pretrained DGM ResNet model checkpoints can be downloaded from the link

After downloading, place the checkpoint files in the same directory as the main.py file.

---

🚀 Training

Stage 1: Geometric Refinement

python main.py --config-name=corgi \
  stage=1 \
  init.prompt="Prompt for Point-E initialization" \
  guidance.control_obj_uid=<Objaverse UID> \
  prompt.prompt="<main prompt>" \
  guidance.guidance_scale=7.5 \
  loss.dgm_step=1 \
  loss.dgm=1e2 \
  guidance.type=controlnet_lora

Stage 2: Texture Refinement

python main.py --config-name=corgi \
  stage=2 \
  init.prompt="Prompt for Point-E initialization" \
  guidance.control_obj_uid=<Objaverse UID> \
  prompt.prompt="<main prompt>" \
  guidance.guidance_scale=7.5 \
  auxiliary.enabled=false \
  renderer.densify.enabled=true \
  renderer.prune.enabled=true \
  loss.dgm_step=1 \
  loss.dgm=1e2 \
  guidance.type=controlnet_lora \
  ckpt="</path/to/stage_1/checkpoint/>"

---

📦 Exporting Outputs

Set the Python path:

export PYTHONPATH="."

Export to .ply

python utils/export.py <your_ckpt> --type ply

Export to .splat

python utils/export.py <your_ckpt> --type splat

Export to Mesh (Shape only)

python utils/export.py <your_ckpt> \
  --type mesh \
  --batch_size 65536 \
  --reso 256 \
  --K 200 \
  --thresh 0.1

Note: <your_ckpt> can be either a .pt checkpoint path or the wandb run ID
(e.g., 0|213630|2023-10-11|a_high_quality_photo_of_a_corgi).
Exported files are saved to exports/<export-type>/.

---