How was the image preprocessed for the model?

[offchan42]

I know that that image is scaled to be in range [-1, 1] when training. But I want to also know whether the images were deformed (resized without caring aspect ratio) during training. And were there any grayscale images in the dataset? The most important thing is about the deformity. My camera has 1280x720 resolution so I want to know if I scale it to 224x224, will I get bad result using transfer learning?