Readme.md
August 14, 2025 · View on GitHub
One View, Many Worlds: Single-Image to 3D Object Meets Generative Domain Randomization for One-Shot 6D Pose Estimation
Zheng Geng, Nan Wang, Shaocong Xu, Bohan Li, Zhaoxi Chen, Chongjie Ye, Sida Peng, Hao Zhao


TODO:
- Release Polished Paper
- Release Evaluation Code
- Release Training Code
⚙️ Installation
Quick Setup Using setup.sh
To streamline the setup process, we provide a setup.sh script that automates all installation steps. Follow these instructions to get started:
Prerequisites
Before running the setup.sh script, ensure you have the following prerequisites installed on your system:
- Python 3.11+
- Conda (recommended)
- Basic build tools (
git,make,cmake, etc.)
Usage
-
Clone the repository:
git clone https://github.com/GZWSAMA/OnePoseviaGen.git cd OnePoseviaGen -
Make the
setup.shscript executable:chmod +x setup.sh -
Run the
setup.shscript:It is recommended to run this script within a fresh conda environment. Here's how you can create and activate a new environment before running the script:
conda create -n OnePoseviaGen python=3.11 -y conda activate OnePoseviaGen ./setup.shThe script will handle:
- Installing PyTorch with CUDA support.
- Installing required dependencies.
- Cloning and installing external extensions.
- Building FoundationPose.
- Installing local packages in editable mode.
- Downloading pretrained weights.
- Patching the transformers library.
-
Verify the setup:
After running the script, verify that all dependencies are correctly installed and the necessary files are downloaded.
-
Web demo:
python app.py
Troubleshooting
Indices should be on the same device
Can't convert cuda:0 device type tensor to numpy
📚 Citation
If you find this work useful, please cite:
TODO