Submitted by Liyii Luo.

Submission data

Full nameWeakly Supervised 3D Scene Parsing with Viewpoint Bottleneck
Publication titlePointly-supervised 3D Scene Parsing with Viewpoint Bottleneck
Publication authorsLiyi Luo, Beiwen Tian, Hao Zhao, Guyue Zhou
Publication URLhttps://arxiv.org/abs/2109.08553
Input Data TypesUses Color,Uses Geometry        Uses 3D
Programming language(s)python with CUDA
HardwareGeForce RTX 3090
Source code or download URLhttps://github.com/OPEN-AIR-SUN/Viewpoint-Bottleneck
Submission creation date4 Jun, 2021
Last edited20 Sep, 2021

3D semantic label results with limited annotations



Infoavg ioubathtubbedbookshelfcabinetchaircountercurtaindeskdoorfloorotherfurniturepicturerefrigeratorshower curtainsinksofatabletoiletwallwindow
0.6690.8470.7320.7240.6130.8270.4430.7420.5620.5510.9470.4410.2180.6500.7530.6210.7650.6010.9050.8140.618