UCSC-VLAA / HQ-Edit

HQ-Edit: A High-Quality and High-Coverage Dataset for General Image Editing

HQ-Edit: A High-Quality and High-Coverage Dataset for General Image Editing

Dataset, code, and model for HQ-Edit.

A working demo with our fine-tuned checkpoint is available.

Check project website for data examples and more.

Dataset Summary

HQ-Edit is a high-quality and high-coverage instruction-based image editing dataset with around 200,000 edits collected with GPT-4V and DALL-E 3. HQ-Edit’s high-resolution images, rich in detail and accompanied by comprehensive editing prompts, substantially enhance the capabilities of existing image editing models.

Create Your Own Dataset

Code Refactoring

Quick Start

Make sure to install the libraries first:

pip install accelerate transformers
pip install git+https://github.com/huggingface/diffusers

import torch
from diffusers import StableDiffusionInstructPix2PixPipeline, EulerAncestralDiscreteScheduler
from diffusers.utils import load_image

image_guidance_scale = 1.5
guidance_scale = 7.0
model_id = "MudeHui/HQ-Edit"
pipe = StableDiffusionInstructPix2PixPipeline.from_pretrained(model_id, torch_dtype=torch.float16, safety_checker=None)
pipe.to("cuda")
pipe.scheduler = EulerAncestralDiscreteScheduler.from_config(pipe.scheduler.config)
resolution = 512
image = load_image(
    "https://hf.co/datasets/diffusers/diffusers-images-docs/resolve/main/mountain.png"
).resize((resolution, resolution))

edit_instruction = "Turn sky into a cloudy one"
edited_image = pipe(
    prompt=edit_instruction,
    image=image,
    height=resolution,
    width=resolution,
    guidance_scale=image_guidance_scale,
    image_guidance_scale=image_guidance_scale,
    num_inference_steps=30,
).images[0]
edited_image.save("edited_image.png")

Citation

If you find our HQ-Edit dataset or the fine-tuned checkpoint useful, please consider citing our paper:

@article{hui2024hq,
  title   = {HQ-Edit: A High-Quality Dataset for Instruction-based Image Editing},
  author  = {Hui, Mude and Yang, Siwei and Zhao, Bingchen and Shi, Yichun and Wang, Heng and Wang, Peng and Zhou, Yuyin and Xie, Cihang},
  journal = {arXiv preprint arXiv:2404.09990},
  year    = {2024}
}