Submitted anonymously.

Submission data

Full nameSSC-UNet64
DescriptionUNet with 64 filters in the first layer, and double the number of filters each time the data is downsampled. Ran with 5cm voxel
https://github.com/facebookresearch/SparseConvNet
Input Data TypesUses Color        Uses 3D
Programming language(s)pytorch
HardwareTitan X
Submission creation date9 Nov, 2018
Last edited10 Nov, 2018

3D semantic label results

Infoavg ioubathtubbedbookshelfcabinetchaircountercurtaindeskdoorfloorotherfurniturepicturerefrigeratorshower curtainsinksofatabletoiletwallwindow
permissive0.3080.3530.2900.2780.1660.5530.1690.2860.1470.1480.9080.1820.0640.0230.0180.3540.3630.3450.5460.6850.278