FASHN VTON v1.5

Run a local try-on inference

Generate and save a try-on image from a person image and a garment image.

Before you start

• Python and a virtual environment
• The repository's model weights, approximately 2 GB, plus human-parser weights downloaded on first use
• A person image and garment image
• CUDA configured for GPU-accelerated pose detection, or CPU ONNX Runtime for CPU-only use

Steps

1. Clone the repository, enter its directory, create and activate a virtual environment, then install the package with the command below.

   Documented commandCopy command

   pip install -e .

2. Download the model and DWPose weights with the command below.

   Documented commandCopy command

   python scripts/download_weights.py --weights-dir ./weights

3. Run the command below with the weights directory, person image, garment image and a category such as tops.

   Documented commandCopy command

   python examples/basic_inference.py

4. Save the returned first image as the output image when using the Python pipeline.

Guide limitations

• Supported categories are tops, bottoms and one-pieces.
• GPU pose detection requires CUDA configuration; the README provides a CPU ONNX Runtime alternative.
• The human-parser weights download automatically on first use.