In the Apple vs Orange code, is it possible to not average the image size and use the original (160x120) size?

[Deleted user]

Thanks for the code but is it possible to use the original size of the image instead of averaged out by block?

[eloquentarduino]

Although technically possible, I cannot see why you would so such a thing, if not using neural networks. Any classifier fed by 160*120 = 19200 features would only get confused.
Moreover, images from the camera suffer from a high degree of noise, so you are strongly recommended to filter that noise out: averaging is one of the possible methods.