Overview

HQ-Edit is developed to support complex instruction-based image editing tasks with about 200,000 edits. This dataset utilizes advanced foundation models to ensure high-quality and aligned diptychs.

Highlights

• Quality controlled image editing prompts and outputs.

• Evaluation metrics for alignment and coherence of edits.

  Significance

  The dataset is a significant step forward in synthetic media, enhancing the capabilities of image editing models, and providing a structured way to leverage modern generative AI models in practical tasks.