Submitted by Xin Lai.

Submission data

Full nameStratified Transformer for 3D Point Cloud Segmentation
DescriptionStratified Transformer adopts a novel stratified sampling strategy, and develops the point embedding module as well as the contextual position encoding to further boost the performance.
Publication titleStratified Transformer for 3D Point Cloud Segmentation
Publication authorsXin Lai*, Jianhui Liu*, Li Jiang, Liwei Wang, Hengshuang Zhao, Shu Liu, Xiaojuan Qi, Jiaya Jia
Publication venueCVPR 2022
Publication URLhttps://arxiv.org/pdf/2203.14508.pdf
Input Data TypesUses Color        Uses 3D
Programming language(s)Python with CUDA
HardwareTesla V100
Source code or download URLhttps://github.com/dvlab-research/Stratified-Transformer
Submission creation date31 Aug, 2021
Last edited20 Apr, 2022

3D semantic label results

Infoavg ioubathtubbedbookshelfcabinetchaircountercurtaindeskdoorfloorotherfurniturepicturerefrigeratorshower curtainsinksofatabletoiletwallwindow
permissive0.7470.9010.8030.8450.7570.8460.5120.8250.6960.6450.9560.5760.2620.7440.8610.7420.7700.7050.8990.8600.734