Is there a way to view output images of the model in PyTorch other than converting to numpy and viewing with matplotlib?
I am training a vggnet-16 model on nyu-v2 depth dataset and I am getting salt and pepper noise in PyTorch whereas I am not getting any salt and pepper noise in Keras.
How are you displaying the images from Keras? Maybe they’re in an incorrect range
For Keras I am using simple Matplotlib functions. The range is same in both frameworks. The problem was with the dataset and the salt and pepper noise diminishes if I increase the number of epochs. It worked in Keras, but haven’t checked in PyTorch yet.