I wanted to implement a pretrained ResNet in my Framework. For that, I want to use the ResNet-18. Since it is not available on Keras I used the one on Github: https://github.com/qubvel/classification_models
I was wondering if the preprocessing of the Input even does anything with that implementation since the preprocessing function for the ResNet models is implemented like this:
def preprocess_input(x, **kwargs): return x
Since I want to use the ResNet to extract features, shouldn't the images need to be preprocessed before I feed them to the network?
2.1m questions
2.1m answers
60 comments
57.0k users