New benchmark and framework tackle complex human-object interaction editing

Researchers have introduced HOI-Edit, a new benchmark designed to evaluate complex Human-Object Interactions (HOI) in image editing, moving beyond static attributes. This benchmark includes an automated metric, HOI-Eval, which assesses instance-level interactions by having vision-language models (VLMs) answer questions after analyzing images with grounded HOI pairs. The study also proposes SCPE (Self-Correcting Process Editing), an agentic framework that refines prompts for Image-to-Video (I2V) models to improve the accuracy of dynamic HOI editing, achieving competitive performance with state-of-the-art models like Nano Banana. AI

IMPACT This research could lead to more sophisticated AI image editing tools capable of understanding and manipulating complex interactions.

RANK_REASON The cluster describes a new academic paper introducing a benchmark and framework for image editing. [lever_c_demoted from research: ic=1 ai=1.0]

Read on arXiv cs.CV →

• HOI-Edit
• HOI-Eval
• Image-to-Video (I2V)
• Jiayi Gao
• Nano Banana
• SCPE (Self-Correcting Process Editing)
• vision-language model

• paper
• product

AI-generated summary · Google Gemini · from 1 sources. How we write summaries →