Submitted anonymously.

Submission data

Full nameSparse Convolutions for Semantic 3D Instance Segmentation
DescriptionSemantic Instance Segmentation based on Mask R-CNN, 3D-SIS and Submanifold Sparse Convolutional Networks
Input Data TypesUses Color,Uses Geometry        Uses 3D
Programming language(s)Python 3.8, PyTorch 1.4
HardwareGeForce GTX 1080 Ti
Submission creation date16 Mar, 2020
Last edited20 May, 2020

3D semantic instance results



Infoavg ap 50%bathtubbedbookshelfcabinetchaircountercurtaindeskdoorotherfurniturepicturerefrigeratorshower curtainsinksofatabletoiletwindow
0.5151.0000.5380.2820.4680.7900.1730.3450.4290.4130.4840.1760.5950.5910.5220.6680.4760.9860.327