Submitted by tong he.

Submission data

Full nameDynamic Convolution
DescriptionDyCo3D: Robust Instance Segmentation of 3D Point Clouds through Dynamic Convolution. CVPR2021
Publication titleDyCo3D: Robust Instance Segmentation of 3D Point Clouds through Dynamic Convolution
Publication authorsTong He; Chunhua Shen; Anton van den Hengel
Publication venueCVPR2021
Publication URLhttps://github.com/aim-uofa/DyCo3D
Input Data TypesUses Color,Uses Geometry        Uses 3D
Programming language(s)C++ with CUDA
Hardware1080ti
Websitehttps://github.com/aim-uofa/DyCo3D
Source code or download URLhttps://github.com/aim-uofa/DyCo3D
Submission creation date10 Nov, 2020
Last edited14 Sep, 2021

3D semantic instance results



Infoavg ap 50%bathtubbedbookshelfcabinetchaircountercurtaindeskdoorotherfurniturepicturerefrigeratorshower curtainsinksofatabletoiletwindow
copyleft0.6411.0000.8410.8930.5310.8020.1150.5880.4480.4380.5370.4300.5500.8570.5340.7640.6570.9870.568