Reproduce “MoDE: Mixture of Diffusion Experts for Any Occluded Face Recognition” 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

Qiannan Fan, Zhuoyang Li, Jitong Li, Chenyang Cao

A framework called MoDE (Mixture of Diffusion Experts) enables robust occluded face recognition by combining multiple diffusion models with an identity-gating network, achieving superior recognition accuracy on both artificial and real-world occluded face datasets while maintaining adaptability through a plug-and-play design.

Key insights

  • Using multiple diffusion experts provides more robust reconstruction than single models
  • Adaptive weighting through ID-Gate is crucial for preserving identity information
  • Iterative resampling helps reduce boundary artifacts between known and reconstructed regions

Summary by alphaXiv.

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