Reproduce “Hierarchical Multimodal Interaction Reasoning for Dyadic Relationship 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
Wang Tang, Fethiye Irmak Dogan, Linbo Qing, Ce Zhu, Hatice Gunes
HiRe-DRR models dyadic relationship recognition by separating low- and high-level cues and homogeneous and cross-modal interactions. A learned router selects relevant interaction tokens while retaining unselected tokens as context, and hierarchical loss supervises categories. On Seamless Interaction, it reaches 63.0% UAR and 68.5% mAP, versus AsyReC’s 50.3% UAR and 52.2% mAP. This result covers 2,738 sampled videos under the reported three-fold evaluation.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.