Instruct-PG: Enhancing Image Editing with Semantic and Preference Alignment

Instruct-PG - Official Implementation

Introduction

This repo, named Instruct-PG, contains the official PyTorch implementation of our paper Instruct-PG: Enhancing Image Editing with Semantic and Preference Alignment. We are actively updating and improving this repository. If you find any bugs or have suggestions, welcome to raise issues or submit pull requests (PR).

Getting Started 🏁

1. Clone the code and prepare the environment

git clone https://github.com/yourusername/Instruct-PG.git
cd Instruct-PG

# create env using conda
conda create -n InstructPG python=3.10
conda activate InstructPG
pip install -r requirements.txt

Nvidia users should install stable pytorch using this command:

pip install torch torchvision torchaudio --extra-index-url https://download.pytorch.org/whl/cu124

2. Download pretrained weights

2.1 Download Stable Diffusion v1.5 weights

# !pip install -U "huggingface_hub[cli]"
huggingface-cli download stable-diffusion-v1-5/stable-diffusion-v1-5

2.2 Download ImageFLow weights

You can download our preference model from here.

3. Train and Inference

python main.py

4. DataSets

You can download our Preference Dataset from here. You can also download our Image Editing Instruction Dataset from here.

Contact

For any queries or collaboration opportunities, please contact us at [email protected].

---

Feel free to explore the code, raise issues, or submit pull requests to enhance the framework further. Happy editing!

Open Source Plan

We are excited to announce that Instruct-PG will be fully open-sourced in the near future! We aim to promote innovation and development within the community by making the source code freely available, thereby advancing image editing technology together.