Reproduce “Semantically Aligned Gradient-Driven Context-Preserving Image Editing” on OpenResearch
Turn your favorite coding agents into research agents that review literature, develop hypotheses, and run experiments.
Universal app for Apple silicon and Intel. Requires macOS 11 or later.
About this paper
Chiranjeev Chiranjeev, Muskan Dosi, Mayank Vatsa, Richa Singh
IABEdit trains image editors to make requested changes in the right regions while preserving context. A frozen vision-language model describes the ground-truth edit, and a trainable aligner compares it with the generated image, sending semantic feedback to the generator; the VLM is not used at inference. On MagicBrush, DINO-I is 88.26, but does not directly measure source-image preservation.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.