如何理解每种模型在keras中的`preprocess_input`函数？

Question

I am trying to understand how the preprocess_input function works in keras for different models. I looked into the source code (ie ResNet50 ) but could not make sense of the decorator @keras_modules_injection which is defined here .

How does module injection work with decorators?

How can I directly see which preprocessings steps are performed for each model?

Answer 1

Actually, the pre-trained models and their relevant codes were factored out a while ago into their own separate package called keras_applications . Therefore, you can see the pre-processing logic, specifically the preprocess_input function here .