Human Object Interaction
Human-object interaction (HOI) research focuses on understanding and modeling how humans interact with objects in images and videos, aiming to accurately detect, classify, and even generate these interactions. Current research emphasizes developing robust models, often leveraging transformer architectures and diffusion models, to handle challenges like occlusion, diverse object categories, and limited training data, particularly in zero-shot and few-shot learning scenarios. This field is crucial for advancing computer vision, robotics, and human-computer interaction, with applications ranging from improved activity recognition and virtual/augmented reality to more intuitive human-robot collaboration and assistive technologies. The development of large-scale, high-quality datasets with detailed annotations is also a significant area of focus.
Papers
FORCE: Dataset and Method for Intuitive Physics Guided Human-object Interaction
Xiaohan Zhang, Bharat Lal Bhatnagar, Sebastian Starke, Ilya Petrov, Vladimir Guzov, Helisa Dhamo, Eduardo Pérez-Pellitero, Gerard Pons-Moll
THOR: Text to Human-Object Interaction Diffusion via Relation Intervention
Qianyang Wu, Ye Shi, Xiaoshui Huang, Jingyi Yu, Lan Xu, Jingya Wang