wkentaro / pytorch-fcn

PyTorch Implementation of Fully Convolutional Networks. (Training code to reproduce the original result is available.)

Final layer deconvolutional filters are not fixed

JJ-res101 opened this issue 4 years ago · comments

JJ-res101 commented 4 years ago

In paper, Final layer deconvolutional filters are fixed to bilinear inertpolation. But I found this implement don't do that.Is that right?

Kentaro Wada commented 3 years ago

It's frozen here

pytorch-fcn/examples/voc/train_fcn32s.py

Line 42 in e86a0f0

# weight is frozen because it is just a bilinear upsampling