In the GoogLeNet implementatoin, the inception class, in branch3 the kernel size of the second convolution is 3. Shouldn’t that be 5 (5X5 convolution after 1X1 used to reduce the tensor depth)?
In the GoogLeNet implementatoin, the inception class, in branch3 the kernel size of the second convolution is 3. Shouldn’t that be 5 (5X5 convolution after 1X1 used to reduce the tensor depth)?