OmniGen2

/

OmniGen2-EditScore7B

like 7

Follow

OmniGen 217

This PR updates the model card to accurately reflect the EditScore model, which is a state-of-the-art reward model for instruction-guided image editing, as presented in the paper "EditScore: Unlocking Online RL for Image Editing via High-Fidelity Reward Modeling".

Key changes include:

• Updated Metadata: Added pipeline_tag: image-to-image, library_name: transformers (due to the use of a Transformers-compatible backbone model like Qwen-VL-2.5), and relevant tags (reward-model, image-editing) to enhance discoverability and proper integration with the Hugging Face ecosystem.
• Comprehensive Content: Replaced the previous OmniGen2-focused content with detailed information about EditScore, including its abstract, key highlights, links to the paper, project page, and GitHub repository, a clear Python usage example, and citation information, all sourced from the official EditScore GitHub repository.
• Visuals: Incorporated relevant images from the EditScore repository to illustrate its capabilities and benchmark results.

This update ensures the model card accurately describes EditScore and provides users with essential information for its understanding and use.