Reproduce “MorphSIG: Subject-Driven Image Generation via Decoupled Anchoring and Feature Transport” 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

Hailong Yan, Yongrui Zhang, Xiangtao Zhang, Le Zhang

MorphSIG addresses subject-driven image generation by preserving a reference subject while changing its pose or scene. It relaxes reference structure early in denoising, then transports subject features into the evolving image using Progressive Decoupled Anchoring and Warping Feature Transport. On ViStoryBench, it raises CIDS-Cross from 0.576 to 0.605 over FLUX.2-4B, though gains are not uniform across metrics.

Summary by alphaXiv.

Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.