Submitted by kangcheng Liu.

Submission data

Full nameFG-Net on Scannet
DescriptionWe present FG-Net, a general deep learning framework for large-scale point clouds understanding without voxelizations, which achieves accurate and real-time performance with a single NVIDIA GTX 1080 8G GPU and an i7 CPU. First, a novel noise and outlier filtering method is designed to facilitate the subsequent high-level understanding tasks. For effective understanding purpose, we propose a novel plug-and-play module consisting of correlated feature mining and deformable convolution based geometric-aware modelling, in which the local feature relationships and point clouds geometric structures can be fully extracted and exploited. For the efficiency issue, we put forward a new composite inverse density sampling based and learning based operation and a feature pyramid based residual learning strategy to save the computational cost and memory consumption respectively.
Publication titlehttps://arxiv.org/abs/2012.09439
Publication authorsKangcheng Liu, Ben M. Chen
Publication venuearXiv Preprint
Publication URLhttps://arxiv.org/abs/2012.09439
Input Data TypesUses Geometry        Uses 3D
Programming language(s)C++
HardwareA Single GTX 1080 GPU
Websitehttps://github.com/KangchengLiu/Feature-Geometric-Net-FG-Net
Source code or download URLhttps://github.com/KangchengLiu/Feature-Geometric-Net-FG-Net
Submission creation date27 Jun, 2021
Last edited14 Sep, 2021

3D semantic label results

Infoavg ioubathtubbedbookshelfcabinetchaircountercurtaindeskdoorfloorotherfurniturepicturerefrigeratorshower curtainsinksofatabletoiletwallwindow
permissive0.6900.8840.7540.7950.6470.8180.4220.8020.6120.6040.9450.4620.1890.5630.8530.7260.7650.6320.9040.8210.606