New method enforces consistent feature transport for image relighting

Researchers have introduced Consistent Feature Transport (CFT), a novel training principle for image relighting that explicitly enforces illumination-consistent transformations between source and target images. By reformulating relighting as an illumination feature transport problem and leveraging rectified flow, CFT jointly models generation and content-preserving transport. This approach aims to isolate illumination variations while maintaining identity and geometry, showing improved results over existing methods and demonstrating potential for other editing tasks like style transfer. AI

IMPACT This research introduces a new technique for image relighting that could improve the quality and consistency of generated images in various editing tasks.

RANK_REASON The cluster contains a research paper detailing a new method for image relighting. [lever_c_demoted from research: ic=1 ai=1.0]

Read on arXiv cs.CV →

• alphaXiv
• arXiv
• CatalyzeX
• Consistent Feature Transport
• DagsHub
• Gotit.pub
• Hugging Face
• Image relighting
• ScienceCast

• paper
• other

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