Official Jittor implementation of "SketchFaceGS: Real-Time Sketch-Driven Face Editing and Generation with Gaussian Splatting".

Tested on Linux with Python 3.10 and an NVIDIA GPU.

git clone https://github.com/YOUR_USERNAME/SketchFaceGS_jittor.git
cd SketchFaceGS_jittor

conda create -n sketchfacegs python=3.10 -y
conda activate sketchfacegs

pip install -r requirements.txt

Then build the bundled JGaussian CUDA ops:

bash scripts/build_jgaussian.sh

Download the main model checkpoint from HuggingFace:

bash scripts/download_weights.sh

This will place the checkpoint at checkpoints/model.pkl. All other required assets are bundled in the repository.

python app_addsketch_jittor.py --checkpoint checkpoints/model.pkl --port 7860

Then open http://127.0.0.1:7860 in your browser.

python infer_jittor.py --checkpoint checkpoints/model.pkl

You can also use the wrapper script:

bash run_inference_jittor.sh

Thanks to these great repositories: JGaussian, GGHead, LHM, Sketch Simplification, and many other inspiring works in the community.