UniHSI: Unified Human-Scene Interaction via Prompted Chain-of-Contacts

Zeqi Xiao1, 2, Tai Wang1, Jingbo Wang1, Jinkun Cao3, Wenwei Zhang1, Bo Dai1, Dahua Lin1 Jiangmiao Pang1*
1Shanghai AI Laboratory, 2Nanyang Technological University, 3Carnegie Mellon University
Paper arXiv Code

UniHSI supports unified and long-horizon control following language commands, enjoying impressive features like fine-granularity control, diverse interactions with the same object, and multi-obj interaction.

Abstract

This paper presents a UNIfied HSI framework, UniHSI, which supports unified control of diverse interactions through language commands. This framework is built upon the definition of interaction as Chain of Contacts (CoC): steps of human joint-object part pairs, which is inspired by the strong correlation between interaction types and human-object contact regions. Based on the definition, UniHSI constitutes a Large Language Model (LLM) Planner to translate language prompts into task plans in the form of CoC, and a Unified Controller that turns CoC into uniform task execution. To facilitate training and evaluation, we collect a new dataset named ScenePlan that encompasses thousands of task plans generated by LLMs based on diverse scenarios. Comprehensive experiments demonstrate the effectiveness of our framework in versatile task execution and generalizability to real scanned scenes.

Framework Pipeline

The whole pipeline consists of two major components: the LLM Planner and the Unified Controller. The LLM planner takes language inputs and background scenario information as inputs and outputs multi-step plans in the form of a Chain of Contacts. The Unified Controller then executes task plans step-by-step and outputs interaction movements.

Multi-objects Interaction

Diverse Interactions with the Same Object

''Multi-agent'' Interaction Planned by LLMs

Using UniHSI, we use LLMs to plan and control multi-agent interaction and synthesis interesting videos. Note: the ''multi-agent interaction'' can be just fulfilled at the commands level at the current stage.

BibTeX


      @inproceedings{
        xiao2024unified,
        title={Unified Human-Scene Interaction via Prompted Chain-of-Contacts},
        author={Zeqi Xiao and Tai Wang and Jingbo Wang and Jinkun Cao and Wenwei Zhang and Bo Dai and Dahua Lin and Jiangmiao Pang},
        booktitle={The Twelfth International Conference on Learning Representations},
        year={2024},
        url={https://openreview.net/forum?id=1vCnDyQkjg}
      }