A research by Mr. Hou et al. was presented at ICCV2025
posted on : October 23, 2025
A research conducted by Trainee Mr. Hou (an internship student last summer), Researcher Dr. Sonokashira, and Team Director Dr. Kawanishi (Multimodal Data Recognition Research Team) has been accepted to the International Conference on Computer Vision (ICCV 2025), one of the top conferences in the field of computer vision, and was presented as a poster.
The study proposes a lightweight method for constructing 3D scene graphs at faster-than-real-time speed.
Details are as follows.
H.-Y. Hou, C.-Y. Lee, M. Sonogashira, Y. Kawanishi
FROSS: Faster-than-Real-Time Online 3D Semantic Scene Graph Generation from RGB-D Images
The 20th International Conference on Computer Vision. 2025/10/23.
https://openaccess.thecvf.com/content/ICCV2025/html/Hou_FROSS_Faster-Than-Real-Time_Online_3D_Semantic_Scene_Graph_Generation_from_RGB-D_ICCV_2025_paper.html
Project page
https://projects.howardkhh.net/FROSS/