Lune

CVPR Findings2026

HypHOI: Exploring Hierarchical Hyperbolic Embeddings for Human-Object Interaction Detection

Yixin Guo, Yu Liu, Weimin Wang, Yanming Guo, Qi Jia

Abstract

Human-object interaction (HOI) detection aims to identify and localize interactions between any human and object, inherently requiring the modeling of complex hierarchical semantic relationships between humans and objects. However, existing methods are predominantly built in Euclidean space, whose polynomial volume growth limits the ability to represent the exponentially expanding hierarchical structures underlying HOIs. This geometric constraint hinders generalization to HOI detection. To overcome this limitation, we propose HypHOI, a hyperbolic representation-based framework for HOI detection. Hy-pHOI exploits the intrinsic advantages of hyperbolic geometry in modeling hierarchical and tree-like structures by embedding HOI triplets into a hyperbolic manifold, thereby preserving hierarchical and scale information with minimal distortion. Furthermore, to mitigate scale inconsistency and better capture hierarchical relationships, we design two core components: a Hyperbolic Adapter to alleviate performance degradation caused by scale variation, and a Hyperbolic Decoder to explicitly model hierarchical dependencies for stronger distinction. Extensive experiments on the HICO-DET and V-COCO benchmarks demonstrate that HypHOI significantly enhances generalization and representation quality, achieving SOTA performance under both fully supervised and zero-shot HOI detection settings.

Ask Lune

Know this paper in minutes, not an afternoon.

Lune can read every page of it and the top-tier papers that cite it, and gives you its claims, its evidence and its limits, each one cited to the paper it comes from. No skimming, no AI slop.

Questions to start from

Lune can draw on

the full paper · related work

Get the answer

Free to start. No credit card required.

Lineage

  • Spatially Conditioned Graphs for Detecting Human-Object Interactions cites HOI Analysis: Integrating and Decomposing Human-Object Interaction
  • Hyperbolic Image-text Representations cites Learning Transferable Visual Models From Natural Language Supervision
Builds onCited byLeft to right by year; the more cited, the higher.

Builds on37

  1. Learning Transferable Visual Models From Natural Language SupervisionAlec Radford, Jong Wook Kim, Chris Hallacy, Aditya Ramesh et al.ICML 2021 · 47,906 citations
  2. Hyperbolic Neural Networks++Ryohei Shimizu, Yusuke Mukuta, Tatsuya HaradaICLR 2021 · 791 citations
  3. Spatially Conditioned Graphs for Detecting Human-Object InteractionsFrederic Z. Zhang, Dylan Campbell, Stephen GouldICCV 2021 · 170 citations
  4. HOI Analysis: Integrating and Decomposing Human-Object InteractionYong-Lu Li, Xinpeng Liu, Xiaoqian Wu, Yizhuo Li et al.NeurIPS 2020 · 152 citations
  5. Hyperbolic Image-text RepresentationsKaran Desai, Maximilian Nickel, Tanmay Rajpurohit, Justin Johnson et al.ICML 2023 · 137 citations

Related papers

  1. Detecting Human-Object Interactions via Functional GeneralizationAnkan Bansal, Sai Saketh Rambhatla, Abhinav Shrivastava, Rama ChellappaAAAI 2020 · 131 citations
  2. End-to-End Human Object Interaction Detection With HOI TransformerCheng Zou, Bohan Wang, Yue Hu, Junqi Liu et al.CVPR 2021
  3. Learning Human-Object Interaction Detection Using Interaction PointsTiancai Wang, Tong Yang, Martin Danelljan, Fahad Shahbaz Khan et al.CVPR 2020
  4. HOTR: End-to-End Human-Object Interaction Detection With TransformersBumsoo Kim, Junhyun Lee, Jaewoo Kang, Eun-Sol Kim et al.CVPR 2021
  5. HORP: Human-Object Relation Priors Guided HOI DetectionPei Geng, Jian Yang, Shanshan ZhangCVPR 2025
Dusk over the sea between two cliffs drawn in fine vertical lines