Name
Abstract | ||
|---|---|---|
Human-object interaction is important information for scene creation and understanding. Most previous studies obtain the interaction-contextual information from the observed data on human-object interaction, but the data collection requires significant amount of time and effort, as well as state-of-the art capturing technique. In addition, the observation-based approach cannot be applied to virtual objects well. As a viable alternative, we propose a novel method to reconstruct synthetic scenes purely from captured motions and to analyze the interaction-contextual information of the synthetic scenes and motions. The scene reconstruction process searches for 3D objects from an object database that match the captured motions, which is achieved by constructing abstract objects containing contact-related information inferred from captured motions. Scene analysis process obtains interaction-contextual information, including interaction behavior, functionality of an object, and the interaction space of an object. We demonstrate the effectiveness of our method through a number of experiments. |
Year | DOI | Venue |
|---|---|---|
2017 | 10.1016/j.gmod.2017.10.002 | Graphical Models |
Keywords | Field | DocType |
Human-object interaction,Scene reconstruction,Scene analysis | Computer vision,Data collection,Scene analysis,Computer graphics (images),Computer science,Scene statistics,Artificial intelligence | Journal |
Volume | ISSN | Citations |
94 | 1524-0703 | 1 |
PageRank | References | Authors |
0.35 | 21 | 2 |
Name | Order | Citations | PageRank |
|---|---|---|---|
| Changgu Kang | 1 | 4 | 0.76 |
| Sung-Hee Lee | 2 | 334 | 24.19 |