Hello Naoki, thanks for your work!
I'm a little confused on the evaluation. I see that the evaluation run is for about 2000 episodes where each time a random target object is selected as the goal. I think you define all the possible objects in habitat_policies.py? (Please correct me if I'm wrong)
I would like to modify the evaluation to only search for one object that I manually specify.
However I cannot understand where this object selection should happen in the code.
Overall, Please help me figure out:
What part of the code selects the target object
How can I specify possible target objects, if I use a different dataset?
Hello Naoki, thanks for your work! I'm a little confused on the evaluation. I see that the evaluation run is for about 2000 episodes where each time a random target object is selected as the goal. I think you define all the possible objects in habitat_policies.py? (Please correct me if I'm wrong)
I would like to modify the evaluation to only search for one object that I manually specify.
However I cannot understand where this object selection should happen in the code. Overall, Please help me figure out: