Input image resize - VGG (224x224)

[sravya tirukkovalur]

Hello,

I am adapting VGG16 model for my image classification problem and my image dimensions are the following:

Mean = [ 399.3507,  374.5622]

Std = [ 107.4896,  100.5482]

I am curious how does keras resize images if my first layer has an input shape of 3x244x244. I am hoping keras upscales/downscales without losing/ distorting features. But would be interested in learning more about how it is done. Any answers or pointers to where can I find more info would be appreciated.

Thanks,

[yohanes...@gmail.com]

afaik keras VGG16 api doesn't handle the resizing so you need to handle it by yourself. Here is one of the example https://gist.github.com/baraldilorenzo/07d7802847aaad0a35d3