Submitted by Lei Yao.

Submission data

Full name3D Scene Instance Segmentation with Superpoint Select and Merge
DescriptionSGIFormer: Semantic-guided and Geometric-enhanced Interleaving Transformer for 3D Instance Segmentation
Publication titleSGIFormer: Semantic-guided and Geometric-enhanced Interleaving Transformer for 3D Instance Segmentation
Publication authorsLei Yao, Yi Wang, Moyun Liu, Lap-Pui Chau
Publication venueTCSVT, 2024
Publication URLhttps://ieeexplore.ieee.org/abstract/document/10753065
Input Data TypesUses Color,Uses Geometry        Uses 3D
Programming language(s)Python
Hardware4090
Source code or download URLhttps://github.com/RayYoh/SGIFormer
Submission creation date7 Feb, 2024
Last edited22 Nov, 2024
Last uploaded9 Jul, 2024

3D semantic instance results



Infoavg ap 50%bathtubbedbookshelfcabinetchaircountercurtaindeskdoorotherfurniturepicturerefrigeratorshower curtainsinksofatabletoiletwindow
0.7991.0000.9150.7100.7290.9250.6640.6700.7700.7660.7390.7370.7001.0000.7920.8290.8150.9970.625