https://github.com/Kotlin/kotlindl/blob/master/docs/loading_trained_model_for_inference.md With reference to this guide, if i were to input a 32x32 image, would the model prediction still work?

Model is trained for a specific input size, so you should resize your image before passing it to a model. You can do it with the image preprocessing dsl like so:

pipeline<BufferedImage>().resize {
    outputWidth = 28
    outputHeight = 28
}.apply(image)

(in this example

image

is a

BufferedImage

you have loaded)

thanks @Julia Beliaeva