How to take the prediction mask of each target in the previous frame?

[longmalongma]

@hkchengrex Because two targets are fixed during training, prev_mask[:,0:1] and prev_mask[:,1:2] can be used to take the masks of the two targets in the previous frame. However, the number of targets during testing is not fixed. How to take the prediction mask of each target in the previous frame?

for i, oi in enumerate(self.enabled_obj):
self.prob[self.enabled_obj,ti-1].cuda()

[hkchengrex]

You're right.
self.enabled_obj is a list containing object indices that we have seen.

[longmalongma]

You're right.
self.enabled_obj is a list containing object indices that we have seen.
@hkchengrex Self. enabled_obj also contains background information?How to extract only a few target prediction masks in self.enabled_obj, because the training code only uses value encoder for each target mask

[hkchengrex]

No. I don't understand what exactly you mean but you can always use indexing to extract whatever you need.
Here is a tutorial: https://numpy.org/doc/stable/reference/arrays.indexing.html

[longmalongma]

No. I don't understand what exactly you mean but you can always use indexing to extract whatever you need.
Here is a tutorial: https://numpy.org/doc/stable/reference/arrays.indexing.html

@hkchengrex Is that how I take the predictive mask of all the targets?
for i, oi in enumerate(self.enabled_obj):
self.prob[self.enabled_obj,ti-1].cuda()
Does self.prob[0,ti-1].cuda() represent background for previous frame?
Does self.prob[1,ti-1].cuda() represent the predictive mask of the first target for previous frame?
Does self.prob[2,ti-1].cuda() represent the predictive mask of the second target for previous frame?

[hkchengrex]

Yes.
That seems to be already explained in another issue that you opened: #60

[longmalongma]

Yes.
That seems to be already explained in another issue that you opened: #60

Ok, thank you very much!

[longmalongma]

Yes.
That seems to be already explained in another issue that you opened: #60

Ok, thank you very much!

@hkchengrex
for i, oi in enumerate(self.enabled_obj):
self.prob[oi, ti-1].cuda().unsqueeze(0))
print(self.prob[oi, ti-1].cuda().size())

             torch.Size([1, 480, 864])
            print(self.prob[self.enabled_obj, ti-1].cuda().size())
             torch.Size([3,1, 480, 864])

So I want to code mask, I have to, right?
self.prob[oi, ti-1].cuda().unsqueeze(0))
or self.prob[oi, ti-1].cuda().unsqueeze(1))

[longmalongma]

Yes.
That seems to be already explained in another issue that you opened: #60

Ok, thank you very much!

@hkchengrex
for i, oi in enumerate(self.enabled_obj):
self.prob[oi, ti-1].cuda().unsqueeze(0))
print(self.prob[oi, ti-1].cuda().size())

             torch.Size([1, 480, 864])
            print(self.prob[self.enabled_obj, ti-1].cuda().size())
             torch.Size([3,1, 480, 864])

So I want to code mask, I have to, right?
self.prob[oi, ti-1].cuda().unsqueeze(0))
or self.prob[oi, ti-1].cuda().unsqueeze(1))

[hkchengrex]

Whether to use unsqueeze depends on your needs.