devendrachaplot / Object-Goal-Navigation

Pytorch code for NeurIPS-20 Paper "Object Goal Navigation using Goal-Oriented Semantic Exploration"
https://devendrachaplot.github.io/projects/semantic-exploration
MIT License
327 stars 60 forks source link

Visualize Ground Truth Semantic Map #15

Closed duxuan321 closed 2 years ago

duxuan321 commented 2 years ago

hi everyone : ) I found the _visualize() in semantic_exp.py is used to render Predicted Semantic Map, and it works pretty well.

But how can we get a Ground Truth Semantic Map like this? It seems that simply replace inputs['sem_map_pred'] with scene_info[floor_idx]['sem_map'](which means ground truth semantic map in objectgoal_env.py) didn't work well.

image
devendrachaplot commented 2 years ago

The ground truth semantic map is stored in the sem_map variable: https://github.com/devendrachaplot/Object-Goal-Navigation/blob/master/envs/habitat/objectgoal_env.py#L112