Submitted anonymously.

Submission data

Full nameA Self Training Approach for Weakly Supervised 3D Semantic Segmentation
DescriptionOnly 0.02% of all the labels are used to train the model
Input Data TypesUses Color,Uses Geometry        Uses 3D
Programming language(s)C++, CUDA, Python
HardwareRTX 2080Ti
Submission creation date13 Mar, 2021
Last edited14 Mar, 2021

3D semantic label results

Infoavg ioubathtubbedbookshelfcabinetchaircountercurtaindeskdoorfloorotherfurniturepicturerefrigeratorshower curtainsinksofatabletoiletwallwindow
0.7010.8250.7960.7230.7160.8320.4330.8160.6340.6090.9690.4180.3440.5590.8330.7150.8080.5600.9020.8470.680