# CoACD Parameter Optimization with RL
This repository uses reinforcement learning to find optimal parameters for CoACD convex decomposition on a given mesh.
## Quick Start
1. **Clone the repository**
```bash
git clone https://github.com/LeonLiu4/coacd.git --recursive
cd coacd-
Create and activate the environment
conda env create -f environment.yml conda activate coacd_env
-
Install PyTorch3D
-
Linux
pip install "git+https://github.com/facebookresearch/pytorch3d.git@stable" -
macOS
export KMP_DUPLICATE_LIB_OK=TRUE pip install --no-build-isolation --no-deps \ git+https://github.com/facebookresearch/pytorch3d.git@stable unset KMP_DUPLICATE_LIB_OK
-
-
Run training
# CPU only CUDA_VISIBLE_DEVICES= python -m coacd_ppo_train --device cpu # GPU python -m coacd_ppo_train