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
Is there a way to segment multiple objects simultaneously on the same output mask for images?
Let's say I want to segment B number of objects in my image. Let's also assume that I have n number of foreground points for each of those B objects. If create a point_coords batch with the shape [B, n, 2], with multimask_output set to False, I get an output of shape [B, 1, img_height, img_width]. Essentially, I get B number of binary masks -- one for each of my B objects.
Is there a way of getting a single non-binary mask, such that each object will be masked with B unique object_id's or class labels?
I saw that this can be done by assigning object_id's to videos, but I was wondering if there's a way of doing this for images.
The text was updated successfully, but these errors were encountered:
Is there a way to segment multiple objects simultaneously on the same output mask for images?
Let's say I want to segment B number of objects in my image. Let's also assume that I have n number of foreground points for each of those B objects. If create a point_coords batch with the shape [B, n, 2], with multimask_output set to False, I get an output of shape [B, 1, img_height, img_width]. Essentially, I get B number of binary masks -- one for each of my B objects.
Is there a way of getting a single non-binary mask, such that each object will be masked with B unique object_id's or class labels?
I saw that this can be done by assigning object_id's to videos, but I was wondering if there's a way of doing this for images.
The text was updated successfully, but these errors were encountered: