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
Hi, I would appreciate it if someone could shed some light on the object pointer feature introduced in SAM2.
As mentioned in the paper: In addition to the spatial memory, we store a list of object pointers as lightweight vectors for high-level semantic information of the object to segment, based on mask decoder output tokens of each frame.
However, in the (inference) code, I just don't see the benefit of this particular token. Moreover, the tracking still seems to work if you disable it. I wonder if someone could share their insights on this.
The text was updated successfully, but these errors were encountered:
I'm now trying to extract the object pointer from the sam2 model and return it.
while I see the num of the object pointer always is 3 while i input 5 points into the sam with multimask_output = True
Have anyone faced this problem before
I've already notice the pointer num was hard coded in the _build_sam_heads, Does it has no realtionship with the decoded mask ?
Hi, I would appreciate it if someone could shed some light on the
object pointer
feature introduced in SAM2.As mentioned in the paper:
In addition to the spatial memory, we store a list of object pointers as lightweight vectors for high-level semantic information of the object to segment, based on mask decoder output tokens of each frame.
However, in the (inference) code, I just don't see the benefit of this particular token. Moreover, the tracking still seems to work if you disable it. I wonder if someone could share their insights on this.
The text was updated successfully, but these errors were encountered: