Reproduce “ALADDIN: Adversarial Latent Diffusion Distillation for One-Step Identity Swapping” 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
Ivan Shchekotov, Timofei Gritsaev, Viacheslav Pirogov, Dmitry Vetrov
ALADDIN adapts a pretrained image generator for face swapping in one network evaluation, transferring a source identity while preserving target attributes. Its student uses teacher-generated pseudo-triplets and adversarial supervision from frozen model features. On 1,000 FFHQ pairs, the λrec=7 student matches APPLE’s identity similarity at 0.40 and has lower pose error, but this comparison is limited to that setting and test.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.