Mapping Semantic Image Pixels to Object Classes in HM3D Semantic Dataset

[sktalha68]

❓ Questions and Help

Hello,

I am using the HM3D semantic dataset and loading the semantic.glb file in Habitat-Sim. I want to accurately map each pixel of the semantic images generated by the sensor to their corresponding object class. A sample of the

Could you provide guidance on how to achieve this? Help from anyone will be much appreciated.

When I run ECCV_2020_Navigation.ipynb I get the following output. Since I have loaded semantic.glb the RGB feed is giving me semantic images. But when I check the unique colors they are too many, and also I do not know the mapping of these colors to their correct object class.