FACTS

Installation

1. Create the environment and install all dependencies:

   # First, create and activate a Python 3.12 virtual environment
   uv venv --python 3.12
   source .venv/bin/activate
   
   # After activating the environment, install all dependencies in one step
   # (Modify the PyTorch-related lines according to your system/CUDA version)
   uv pip install \
       torch==2.5.1 \
       torchvision==0.20.1 \
       torchaudio==2.5.1 \
       -r requirements.txt

   Note: On Windows, the activation command is .venv\Scripts\activate.

2. Install GroundedSAM:

   .venv\Scripts\activate
   mkdir -p modules
   cd modules
   git clone https://github.com/SoluteToNight/GroundedSam4FACT.git
   cd GroundedSam4FACT
   uv pip install -e .
   uv pip install --no-build-isolation -e grounding_dino

   For more details, see the Grounded-SAM-2 repository.

Usage

The main script for running the processing workflow is demo.py.

1. Place your 3D model assets (e.g., .obj, .mtl, and texture images) into a subdirectory inside the obj folder. See obj/example1 for reference.

2. Run the script from the command line. You can specify the input directory for your model and where to save the output.

   python demo.py --input ./obj/example1 --output ./outputs/example1

Command-Line Arguments

• --input: Path to the directory containing the input model and textures.
• --output: Directory where processed textures will be saved.
• --device: Computation device, cuda or cpu (Default: cuda).