real-stanford / sscnav Goto Github PK

Hello, does your experiment compare with SemExp（Object Goal Navigation using Goal-Oriented Semantic Exploration）?

hello，could you share the dataset about "../buffer/“ directory？I can‘t find it in your download link，thank you

Hi, thank you for providing the nice implementation of your approach.

I ran the test_agent.py for testing the Q agent with ACNet semantic segmentation. There are however certain issues which I am facing:

1. test_agent.py throws error in line 242 when drawing circle because agent.action is a tuple of Tensor. I believe this is related to the action_picker function in SCNav_agent where greedy action gives a tuple of Tensor but the random action returns a tuple of int.
2. I temporarily circumvent the above error by converting to int before passing to draw the circle. Though for the val.json - I get an error Math::Quaternion::fromMatrix(): the matrix is not orthogonal: after the 13/500 step in the first episode(zsNo4HB9uLZ). Do you have any leads for what might be wrong?

Thank you for your help.

Thanks for your excellent work. I'm curious how can I get visualizations of semantic maps and trajectories? Just like the qualitative analysis in the paper. Looking forward to your reply, thank you very much.