How to use vgg19 for a segmentation task? The output layer would be just be 2D tensor from vgg whereas my input has 4 channels which is RGB and an extra channel.
Th error which is thrown currently is
NotImplementedError: Input Error: Only 3D, 4D and 5D input Tensors supported (got 2D) for the modes: nearest | linear | bilinear | trilinear (got UpsamplingNearest2d)
This is because of the difference in shape of the tensor between output from vgg and the input to the network. How to handle the above situation.