Submitted by Wang Luowei.

Submission data

Full nameFusing self-supervised depth completion with residual encoder-decoder network
DescriptionThis is a unified framework incorporating self-supervised depth completion with the residual encoder-decoder segmentation, proposed by AI department, CloudMinds Inc (CM-CV&AR).
Input Data TypesUses Color,Uses Geometry        Uses 2D
Programming language(s)PyTorch
HardwareGeForce 1080Ti 12GB,i7-8700k
Submission creation date5 Mar, 2019
Last edited6 Jun, 2019

2D semantic label results

Infoavg ioubathtubbedbookshelfcabinetchaircountercurtaindeskdoorfloorotherfurniturepicturerefrigeratorshower curtainsinksofatabletoiletwallwindow
0.5830.6820.7230.5420.5100.3100.4510.6680.5490.5200.9200.3750.4460.5280.4170.6700.5770.4780.8620.8060.628