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
It'd be helpful for r to take on input's dimensionality - i.e. measure receptive field of height and width separately, in case strides and kernel sizes aren't equal throughout the network. The current workaround is to move the dimension of interest to be the first - so if (width, height) = (100, 200), we do (200, 100) and swap all network parameters.
The text was updated successfully, but these errors were encountered:
Good point, this change permeates through various parts of the library, so there is no quick fix to change this.
That being said, I see your point and I will come back with some estimates on how long it will take to implement.
Hi,
I managed to build a little prototype, there also will be a new color.
Yellow, which is in cases where the receptive field is on some axis bigger than the input resolution bot not all axes.
This happens for example in InceptionV3, which makes some use of non-square receptive field sizes. inceptionv3_32_pixels.pdf
Thanks for your work.
It'd be helpful for
r
to take on input's dimensionality - i.e. measure receptive field of height and width separately, in case strides and kernel sizes aren't equal throughout the network. The current workaround is to move the dimension of interest to be the first - so if(width, height) = (100, 200)
, we do(200, 100)
and swap all network parameters.The text was updated successfully, but these errors were encountered: