Submitted by Pyunghwan Ahn.

Submission data

Full nameProjection-based Point Convolutional Network
DescriptionThe convolutional module used in this method consists of MLPs and projection-based 2D convolutional modules.
Publication titleProjection-based Point Convolution for Efficient Point Cloud Segmentation
Publication authorsPyunghwan Ahn, Juyoung Yang, Eojindl Yi, Chanho Lee, Junmo Kim
Publication venueIEEE Access
Publication URLhttps://ieeexplore.ieee.org/document/9687584
Input Data TypesUses Color        Uses 3D
Programming language(s)Python(PyTorch) with CUDA
HardwareSilver 4210, RTX 2080Ti, 128GB RAM
Source code or download URLhttps://github.com/pahn04/PPConv
Submission creation date1 Jan, 2022
Last edited1 Apr, 2022

3D semantic label results

Infoavg ioubathtubbedbookshelfcabinetchaircountercurtaindeskdoorfloorotherfurniturepicturerefrigeratorshower curtainsinksofatabletoiletwallwindow
permissive0.6630.7460.7080.7220.6380.8200.4510.5660.5990.5410.9500.5100.3130.6480.8190.6160.6820.5900.8690.8100.656