Software & Data Downloads — MMHOI

MMHOI Dataset: Modeling Complex 3D Multi-Human Multi-Object Interactions for fostering research in multi-human multi-object interactions and serving as a benchmark for next-generation HOI algorithms.

The MMHOI dataset addresses the challenge of modeling real-world scenes with multiple humans and objects interacting in causal, goal-oriented, or cooperative ways. Existing 3D human-object interaction (HOI) benchmarks capture only a subset of these complex interactions. MMHOI closes this gap by providing a large-scale dataset with comprehensive 3D shape and pose annotations for every person and object, covering 78 action categories and 14 interaction-specific body parts across 12 everyday scenarios.

MMHOI is designed to foster research in multi-human multi-object interactions and serves as a benchmark for next-generation HOI algorithms. All necessary data for training, validation, and testing are publicly released.

    •  Kogashi, K., Cherian, A., Kuo, M.-Y.J., "MMHOI: Modeling Complex 3D Multi-Human Multi-Object Interactions", IEEE Winter Conference on Applications of Computer Vision (WACV), March 2026.
      BibTeX TR2026-029 PDF Video Data
      • @inproceedings{Kogashi2026mar,
      • author = {Kogashi, Kaen and Cherian, Anoop and Kuo, Meng-Yu Jennifer},
      • title = {{MMHOI: Modeling Complex 3D Multi-Human Multi-Object Interactions}},
      • booktitle = {IEEE Winter Conference on Applications of Computer Vision (WACV)},
      • year = 2026,
      • month = mar,
      • url = {https://www.merl.com/publications/TR2026-029}
      • }

    Access data at https://doi.org/10.5281/zenodo.17711785.