You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
Hello, I am currently trying to deploy PoliFormer in another simulator, and I have obtained the target object’s bbox in the format of (x, y, w, h). However, I am encountering difficulties when inputting this bbox into the model. I would like to ask how PoliFormer defines its bbox?
Additionally, I have an extra question: what is the difference between nav_task_relevant_object_bbox and nav_accurate_object_bbox, and how are they each obtained?
Thank you for your response.
The text was updated successfully, but these errors were encountered:
Hi!
The bounding box is defined by [x1, y1, x2, y2, area] and those values would be 1000 if no object is detected. Here is the place we convert box info to the format.
The difference between nav_task_relevant_object_bbox and nav_accurate_object_bbox is that nav_task_relevant_object_bbox gets bounding box from segmentation mask from the environment (get_approx_object_mask), while nav_accurate_object_bbox gets the bounding box directly from the simulation engine.
They should perform similar, but nav_task_relevant_object_bbox should run much quicker!
Hello, I am currently trying to deploy PoliFormer in another simulator, and I have obtained the target object’s bbox in the format of (x, y, w, h). However, I am encountering difficulties when inputting this bbox into the model. I would like to ask how PoliFormer defines its bbox?
Additionally, I have an extra question: what is the difference between
nav_task_relevant_object_bbox
andnav_accurate_object_bbox
, and how are they each obtained?Thank you for your response.
The text was updated successfully, but these errors were encountered: