Summary of Himo: a New Benchmark For Full-body Human Interacting with Multiple Objects, by Xintao Lv et al.
HIMO: A New Benchmark for Full-Body Human Interacting with Multiple Objects
by Xintao Lv, Liang Xu, Yichao Yan, Xin Jin, Congsheng Xu, Shuwen Wu, Yifan Liu, Lincheng Li, Mengxiao Bi, Wenjun Zeng, Xiaokang Yang
First submitted to arxiv on: 17 Jul 2024
Categories
- Main: Computer Vision and Pattern Recognition (cs.CV)
- Secondary: Artificial Intelligence (cs.AI)
GrooveSquid.com Paper Summaries
GrooveSquid.com’s goal is to make artificial intelligence research accessible by summarizing AI papers in simpler terms. Each summary below covers the same AI paper, written at different levels of difficulty. The medium difficulty and low difficulty versions are original summaries written by GrooveSquid.com, while the high difficulty version is the paper’s original abstract. Feel free to learn from the version that suits you best!
| Summary difficulty | Written by | Summary |
|---|---|---|
| High | Paper authors | High Difficulty Summary Read the original abstract here |
| Medium | GrooveSquid.com (original content) | Medium Difficulty Summary The paper proposes HIMO, a large-scale dataset of full-body human-object interactions (HOIs) that includes 3.3K 4D sequences and 4.08M 3D frames. Unlike existing datasets, which typically focus on single objects, HIMO captures the manipulation of multiple objects. The authors also introduce two novel tasks for HOI synthesis: conditioning on whole text prompts or segmented text prompts for fine-grained timeline control. To address these tasks, they design a dual-branch conditional diffusion model with a mutual interaction module and an auto-regressive generation pipeline. Experimental results demonstrate the dataset’s generalization ability to unseen object geometries and temporal compositions. |
| Low | GrooveSquid.com (original content) | Low Difficulty Summary This paper creates a big database of people interacting with multiple objects. Right now, most datasets only show people using one object at a time. The new database has 3,300 sequences and 4 million frames of people doing different things with many objects. The researchers also came up with two new ways to make fake HOI videos based on text prompts. They used a special computer model to do this. When they tested the model, it was good at making new videos that looked like they were part of the original database. |
Keywords
» Artificial intelligence » Diffusion model » Generalization
