Thanks for your great work! but I found a problem below...
in your code _super-events-cvpr18/temporal_structurefilter.py line 41-46
def get_filters(self, delta, gamma, center, length, time):
...
a = Variable(torch.zeros(self.Ni))
a = a.cuda()
# stride and center
a = deltas[:, None] * a[None, :]
a = centers[:, None] + a
why always multiply deltas to zeros? do you need to exchange line 45 and line 46
Thanks for your great work! but I found a problem below...
in your code _super-events-cvpr18/temporal_structurefilter.py line 41-46
why always multiply deltas to zeros? do you need to exchange line 45 and line 46