Submitted by gengxin liu.

Submission data

Full nameActive Self-Training for Weakly Supervised 3D Scene Semantic Segmentation
DescriptionActive learning with Self-training
Publication titleActive Self-Training for Weakly Supervised 3D Scene Semantic Segmentation
Publication authorsGengxin Liu, Oliver van Kaick, Hui Huang, Ruizhen Hu
Publication URLhttps://arxiv.org/abs/2209.07069
Input Data TypesUses Color,Uses Geometry        Uses 3D
Programming language(s)python
HardwareRTX 3090
Websitehttps://arxiv.org/abs/2209.07069
Source code or download URLhttps://github.com/GengxinLiu/ActiveST
Submission creation date14 Nov, 2021
Last edited4 Oct, 2022

3D semantic label results with limited annotations



Infoavg ioubathtubbedbookshelfcabinetchaircountercurtaindeskdoorfloorotherfurniturepicturerefrigeratorshower curtainsinksofatabletoiletwallwindow
0.7480.9840.8040.7590.7200.8490.5160.7910.6700.6540.9740.4950.3820.8110.8280.7870.7800.6400.9520.8610.701