Source: Deep Learning on Medium
Thank you for the article. This part which I highlighted confused me a bit. It indicates that when using transfer learning with a pre-trained model, one should check the parameters used in the original training, published in the official paper. However I checked several sources including the documentation I think these parameters are set up automatically by both Tensorflow and Pytorch with the models. The only thing to care about is to transform the pixel values into the range of (0,1). I would appreciate if you could confirm me.