Submitted by Martin Bokeloh.

Submission data

Full name3DMV with Deeplab and FTSDF
DescriptionReimplementation of "3DMV: Joint 3D-Multi-View Prediction for 3D Semantic Scene Segmentation, Angela Dai, Matthias Nießner" with some modifications:

1. We use Deeplab instead of ENet
2. We only fine-tune Deeplab on the 2D labels without end-to-end training
3. As geometry channel we only use the flipped TSDF described in
Semantic Scene Completion from a Single Depth Image
Shuran Song Fisher Yu Andy Zeng Angel X. Chang Manolis Savva Thomas Funkhouser
Publication authorsMartin Bokeloh, Juergen Sturm
Input Data TypesUses Color,Uses Geometry        Uses 2D,Uses 3D
Programming language(s)C++/Python/Tensorflow
HardwareV100
Submission creation date2 Oct, 2018
Last edited2 Oct, 2018

3D semantic label results

Infoavg ioubathtubbedbookshelfcabinetchaircountercurtaindeskdoorfloorotherfurniturepicturerefrigeratorshower curtainsinksofatabletoiletwallwindow
0.5010.5580.6080.4240.4780.6900.2460.5860.4680.4500.9110.3940.1600.4380.2120.4320.5410.4750.7420.7270.477