the input size of image is fixed?

[InkdyeHuang]

[GreatNewHope]

I am pretty sure it is not. I mean, you can train the cnn with almost any size that you want, but it has to be the same throught all training, validation and testing. It is important to remember that images have to be of a minimum size depending on the model (inception minimum is larger, vgg16 is lower). Refer to the original documentation of the models to read more about this.