preprocess and postprocess

[gitlabspy]

Let's say we have an arbitrary size image, it should be preprocessed before feeding into the compression model. When decompression, we retrieve the bitstream and reconstruct the image, then we do postprocess to retrieve the image with its original size.
If we decided the model input shape as [128, 128] and we have two images with shape [256, 128] and [64, 64], how will they be preprocessed and postprocessed?

[jonarchist]

All the pre-trained models accessible with tfci.py and the model code under models/ are fully convolutional. So the decoder will always output an image of the original size.