LOCATEdit: Graph Laplacian Optimized Cross Attention for Localized Text-Guided Image Editing

🚀 Getting Started

Environment Requirement 🌍

conda create -n locatedit python=3.9
conda activate locatedit
pip install -r requirements.txt

🏃🏼 Running Scripts

Inference and Evaluation📜

Run the Benchmark

Run LOCATEdit on PIE-Bench:

python run_editing_locatedit.py --anno_file data/mapping_file.json --image_dir data/annotation_images --sd_model_dir runwayml/stable-diffusion-v1-5 --ip_adapter_dir h94/IP-Adapter --clip_model_dir laion/CLIP-ViT-H-14-laion2B-s32B-b79K

You can specify --output_dir for output path.

Run Any Image

You can process your own images and editing prompts.

python run_editing_locatedit_one_image.py --image_path /path/to/source/image --source_prompt "source prompt" --target_prompt "target prompt" --sd_model_dir runwayml/stable-diffusion-v1-5 --ip_adapter_dir h94/IP-Adapter --clip_model_dir laion/CLIP-ViT-H-14-laion2B-s32B-b79K

You can specify --output_path for output path.

💖 Acknowledgement

Our code is modified on the basis of PnP Inversion, InfEdit and P2P.